Populated PAP

This study is an expansion of an earlier pilot study by the authors, of which the results were used to generate the secondary hypotheses and exploratory analysis plan, in addition to the defined planned contrasts. All primary hypotheses remain identical to that of the pilot study, with the inclusion of hypotheses for the Democratic portion of the sample, as the pilot study sampled solely Republicans. All participants will be newly sampled from Prolific so as to avoid repeated exposure to the treatment.

Descriptive Analyses

I will perform a set of descriptive analyses intended to measure the exploratory relationship between each of our identified measures and condition assignment. I create one table comparing Democrats in the control condition to Democrats in the treatment condition on the basis of the mean and standard deviations for the following variables: partisan self-concept, self-reported ideology, the difference in feeling thermometer ratings between pre- and post-treatment for both parties, the difference between these affective differences, moral disengagement toward Republicans, individual levels of SPV, and meta-perceptions of Republicans’ SPV.

# 4.1: make groups
gD_control <- 2
gD_treat <- 4
D_control <- descriptive_stats %>% filter(exp_group_numeric == gD_control)
D_treat <- descriptive_stats %>% filter(exp_group_numeric == gD_treat)
# 4.2: form column structure
dem_rows <- lapply(names(items), function(label) {
  cols    <- items[[label]]
  mean_col <- cols[1]; sd_col <- cols[2]
# 4.3: how to pull means and SDs for each column         
  valD_control <- if (!is.na(D_control[[mean_col]])) {
    if (mean_col == "count") sprintf("%d", D_control[[mean_col]]) else sprintf("%.2f (%.2f)", D_control[[mean_col]], D_control[[sd_col]])
  } else NA_character_
  star <- if (label %in% names(stars_D)) stars_D[[label]] else ""
  valD_treat <- if (!is.na(D_treat[[mean_col]])) {
    if (mean_col == "count") sprintf("%d", D_treat[[mean_col]])
    else sprintf("%.2f (%.2f)%s", D_treat[[mean_col]], D_treat[[sd_col]], star)
  } else NA_character_
# 4.4: bind into table         
  tibble::tibble(Variable = label, Group_D_control = valD_control, Group_D_treat = valD_treat)
}) %>% dplyr::bind_rows()
# 4.5: make the labels for table and fill in      
nD_control <- D_control %>% pull(count) %>% unique()
nD_treat <- D_treat %>% pull(count) %>% unique()
colnames(dem_rows)[2:3] <- c(
  paste0("Control x Democrat"),
  paste0("Treatment x Democrat"))

Variable	Control x Democrat	Treatment x Democrat
Partisan Self-Concept	3.09 (1.02)	3.13 (1.02)
Self-Reported Ideology	2.24 (1.17)	2.00 (1.00)*
Difference in Affective Ratings:
Republican Party	-0.18 (5.98)	-2.20 (6.06)***
Difference in Affective Ratings:
Democratic Party	-0.08 (5.61)	0.44 (7.93)
Difference in Affective Ratings Between Parties	-0.10 (8.11)	-2.93 (8.55)***
Moral Disengagement: Opposing Party	3.35 (0.98)	3.46 (1.07)
Individual SPV	21.91 (18.20)	25.79 (20.67)*
Meta-Perceptions of Republican SPV	41.77 (25.58)	47.50 (27.94)*
N	251	278

I will create a second table comparing Republicans in the control condition to Republicans in the treatment condition across the same set of variables.

# 3.1: make groups
gR_control <- 2
gR_treat <- 4
R_control <- descriptive_stats %>% filter(exp_group_numeric == gR_control)
R_treat <- descriptive_stats %>% filter(exp_group_numeric == gR_treat)
# 3.2: form column structure
rep_rows <- lapply(names(items), function(label) {
  cols    <- items[[label]]
  mean_col <- cols[1]; sd_col <- cols[2]
# 3.3: how to pull means and SRs for each column         
  valR_control <- if (!is.na(R_control[[mean_col]])) {
    if (mean_col == "count") sprintf("%d", R_control[[mean_col]]) else sprintf("%.2f (%.2f)", R_control[[mean_col]], R_control[[sd_col]])
  } else NA_character_
  star <- if (label %in% names(stars_R)) stars_R[[label]] else ""
  valR_treat <- if (!is.na(R_treat[[mean_col]])) {
    if (mean_col == "count") sprintf("%d", R_treat[[mean_col]])
    else sprintf("%.2f (%.2f)%s", R_treat[[mean_col]], R_treat[[sd_col]], star)
  } else NA_character_
# 3.4: bind into table         
  tibble::tibble(Variable = label, Group_R_control = valR_control, Group_R_treat = valR_treat)
}) %>% dplyr::bind_rows()
# 3.5: make the labels for table and fill in      
nR_control <- R_control %>% pull(count) %>% unique()
nR_treat <- R_treat %>% pull(count) %>% unique()
colnames(rep_rows)[2:3] <- c(
  paste0("Control x Republican"),
  paste0("Treatment x Republican"))

Variable	Control x Republican	Treatment x Republican
Partisan Self-Concept	3.09 (1.02)	3.13 (1.02)
Self-Reported Ideology	2.24 (1.17)	2.00 (1.00)
Difference in Affective Ratings:
Republican Party	-0.18 (5.98)	-2.20 (6.06)
Difference in Affective Ratings:
Democratic Party	-0.08 (5.61)	0.44 (7.93)
Difference in Affective Ratings Between Parties	-0.10 (8.11)	-2.93 (8.55)
Moral Disengagement: Opposing Party	3.35 (0.98)	3.46 (1.07)
Individual SPV	21.91 (18.20)	25.79 (20.67)
Meta-Perceptions of Republican SPV	41.77 (25.58)	47.50 (27.94)
N	251	278

I also created a combined version of this table with four columns by condition assignment and (non-randomized) partisan ID, which was not specified in the pre-analysis plan, but essentially places these two tables next to each other. Therefore, this is not a significant deviation.

# 5.1: make groups
groups <- list(
  list(num = 1, col = "Control x Republican"),
  list(num = 3, col = "Treatment x Republican"),
  list(num = 2, col = "Control x Democrat"),
  list(num = 4, col = "Treatment x Democrat"))
# 5.2: form column structure   
combined_rows <- lapply(names(items), function(label) {
  cols     <- items[[label]]
  mean_col <- cols[1]; sd_col <- cols[2]
# 5.3: how to pull means and SDs for each column  
  vals <- lapply(groups, function(g) {
    grp <- descriptive_stats %>% filter(exp_group_numeric == g$num)
    if (!is.na(grp[[mean_col]])) {
      if (mean_col == "count") sprintf("%d", grp[[mean_col]]) 
      else sprintf("%.2f (%.2f)", grp[[mean_col]], grp[[sd_col]])
    } else NA_character_
  })
# 5.4: bind into table
  row <- tibble::tibble(Variable = label)
  for (i in seq_along(groups)) row[[groups[[i]]$col]] <- vals[[i]]
  row
}) %>% dplyr::bind_rows()

Variable	Control x Republican	Treatment x Republican	Control x Democrat	Treatment x Democrat
Partisan Self-Concept	3.14 (1.05)	3.08 (1.11)	3.09 (1.02)	3.13 (1.02)
Self-Reported Ideology	5.53 (1.08)	5.57 (1.20)	2.24 (1.17)	2.00 (1.00)
Difference in Affective Ratings:
Republican Party	-0.25 (5.95)	-1.90 (9.29)	-0.18 (5.98)	-2.20 (6.06)
Difference in Affective Ratings:
Democratic Party	-0.18 (6.28)	-0.67 (9.72)	-0.08 (5.61)	0.44 (7.93)
Difference in Affective Ratings Between Parties	-0.08 (9.35)	-1.20 (13.19)	-0.10 (8.11)	-2.93 (8.55)
Moral Disengagement: Opposing Party	2.80 (1.23)	2.79 (1.18)	3.35 (0.98)	3.46 (1.07)
Individual SPV	20.50 (21.25)	18.53 (21.26)	21.91 (18.20)	25.79 (20.67)
Meta-Perceptions of Republican SPV	29.95 (19.20)	32.17 (24.24)	41.77 (25.58)	47.50 (27.94)
N	146	114	251	278

I deviate from the pre-analysis plan by creating an additional table that reports the mean and standard errors for each condition. This was selected for inclusion in the final manuscript to be consistent with later analyses, given that our conditional average treatment effect (CATE) estimates report robust standard errors.

Republicans

# 3.1: make groups
gR_control <- 1
gR_treat   <- 3
R_control <- descriptive_stats %>% filter(exp_group_numeric == gR_control)
R_treat <- descriptive_stats %>% filter(exp_group_numeric == gR_treat)
# 3.2: form column structure
rep_rows_se <- lapply(names(items), function(label) {
cols    <- items[[label]]
mean_col <- cols[1]; se_col <- cols[2]
# 3.3: how to pull means and SRs for each column
valR_control_se <- if (!is.na(R_control[[mean_col]])) {
if (mean_col == "count") sprintf("%d", R_control[[mean_col]]) else sprintf("%.2f (%.2f)", R_control[[mean_col]], R_control[[se_col]])
} else NA_character_
star <- if (label %in% names(stars_R)) stars_R[[label]] else ""
valR_treat_se <- if (!is.na(R_treat[[mean_col]])) {
if (mean_col == "count") sprintf("%d", R_treat[[mean_col]])
else sprintf("%.2f (%.2f)%s", R_treat[[mean_col]], R_treat[[se_col]], star)
} else NA_character_
# 3.4: bind into table
tibble::tibble(Variable = label, Group_R_control = valR_control_se, Group_R_treat = valR_treat_se)
}) %>% dplyr::bind_rows()
# 3.5: make the labels for table and fill in
nR_control_se <- R_control %>% pull(count) %>% unique()
nR_treat_se <- R_treat %>% pull(count) %>% unique()
colnames(rep_rows_se)[2:3] <- c(
paste0("Control x Republican"),
paste0("Treatment x Republican"))

Variable	Control x Republican	Treatment x Republican
Partisan Self-Concept	3.14 (1.05)	3.08 (1.11)
Self-Reported Ideology	5.53 (1.08)	5.57 (1.20)
Difference in Affective Ratings:
Republican Party	-0.25 (5.95)	-1.90 (9.29)
Difference in Affective Ratings:
Democratic Party	-0.18 (6.28)	-0.67 (9.72)
Difference in Affective Ratings Between Parties	-0.08 (9.35)	-1.20 (13.19)
Moral Disengagement: Opposing Party	2.80 (1.23)	2.79 (1.18)
Individual SPV	20.50 (21.25)	18.53 (21.26)
Meta-Perceptions of Republican SPV	29.95 (19.20)	32.17 (24.24)
N	146	114

Democrats

# 4.1: make groups
gD_control <- 2
gD_treat <- 4
D_control <- descriptive_stats %>% filter(exp_group_numeric == gD_control)
D_treat <- descriptive_stats %>% filter(exp_group_numeric == gD_treat)
# 4.2: form column structure
dem_rows_se <- lapply(names(items), function(label) {
        cols    <- items[[label]]
        mean_col <- cols[1]; se_col <- cols[2]
# 4.3: how to pull means and SDs for each column
        valD_control_se <- if (!is.na(D_control[[mean_col]])) {
        if (mean_col == "count") sprintf("%d", D_control[[mean_col]]) else sprintf("%.2f (%.2f)", D_control[[mean_col]], D_control[[se_col]])
        } else NA_character_
        star <- if (label %in% names(stars_D)) stars_D[[label]] else ""
        valD_treat_se <- if (!is.na(D_treat[[mean_col]])) {
        if (mean_col == "count") sprintf("%d", D_treat[[mean_col]])
        else sprintf("%.2f (%.2f)%s", D_treat[[mean_col]], D_treat[[se_col]], star)
        } else NA_character_
# 4.4: bind into table
tibble::tibble(Variable = label, Group_D_control = valD_control_se, Group_D_treat = valD_treat_se)
}) %>% dplyr::bind_rows()
# 4.5: make the labels for table and fill in
nD_control_se <- D_control %>% pull(count) %>% unique()
nD_treat_se <- D_treat %>% pull(count) %>% unique()
colnames(dem_rows_se)[2:3] <- c(
paste0("Control x Democrat"),
paste0("Treatment x Democrat"))

Variable	Control x Democrat	Treatment x Democrat
Partisan Self-Concept	3.09 (1.02)	3.13 (1.02)
Self-Reported Ideology	2.24 (1.17)	2.00 (1.00)*
Difference in Affective Ratings:
Republican Party	-0.18 (5.98)	-2.20 (6.06)***
Difference in Affective Ratings:
Democratic Party	-0.08 (5.61)	0.44 (7.93)
Difference in Affective Ratings Between Parties	-0.10 (8.11)	-2.93 (8.55)***
Moral Disengagement: Opposing Party	3.35 (0.98)	3.46 (1.07)
Individual SPV	21.91 (18.20)	25.79 (20.67)*
Meta-Perceptions of Republican SPV	41.77 (25.58)	47.50 (27.94)*
N	251	278

Combined

groups_se <- list(
list(num = 1, col = "Control x Republican"),
list(num = 3, col = "Treatment x Republican"),
list(num = 2, col = "Control x Democrat"),
list(num = 4, col = "Treatment x Democrat"))
# 5.2: form column structure
combined_rows_se <- lapply(names(items), function(label) {
cols     <- items[[label]]
mean_col <- cols[1]; se_col <- cols[2]
# 5.3: how to pull means and SDs for each column
vals_se <- lapply(groups, function(g) {
grp <- descriptive_stats %>% filter(exp_group_numeric == g$num)
if (!is.na(grp[[mean_col]])) {
if (mean_col == "count") sprintf("%d", grp[[mean_col]])
else sprintf("%.2f (%.2f)", grp[[mean_col]], grp[[se_col]])
} else NA_character_
})
# 5.4: bind into table
row <- tibble::tibble(Variable = label)
for (i in seq_along(groups)) row[[groups[[i]]$col]] <- vals_se[[i]]
row
}) %>% dplyr::bind_rows()

Variable	Control x Republican	Treatment x Republican	Control x Democrat	Treatment x Democrat
Partisan Self-Concept	3.14 (1.05)	3.08 (1.11)	3.09 (1.02)	3.13 (1.02)
Self-Reported Ideology	5.53 (1.08)	5.57 (1.20)	2.24 (1.17)	2.00 (1.00)
Difference in Affective Ratings:
Republican Party	-0.25 (5.95)	-1.90 (9.29)	-0.18 (5.98)	-2.20 (6.06)
Difference in Affective Ratings:
Democratic Party	-0.18 (6.28)	-0.67 (9.72)	-0.08 (5.61)	0.44 (7.93)
Difference in Affective Ratings Between Parties	-0.08 (9.35)	-1.20 (13.19)	-0.10 (8.11)	-2.93 (8.55)
Moral Disengagement: Opposing Party	2.80 (1.23)	2.79 (1.18)	3.35 (0.98)	3.46 (1.07)
Individual SPV	20.50 (21.25)	18.53 (21.26)	21.91 (18.20)	25.79 (20.67)
Meta-Perceptions of Republican SPV	29.95 (19.20)	32.17 (24.24)	41.77 (25.58)	47.50 (27.94)
N	146	114	251	278

Correlations

Next, I will create a correlation matrix to test the relationship between individual SPV, meta-perceptions of Republicans’ SPV, and the difference between pre- and post-treatment affective ratings based on whether the party is one’s own party (ingroup) or the opposing party (outgroup).

# A: SPV v. in-party meta-SPV
cor.test(data_clean$SPV_individual, data_clean$SPV_meta_R)

## 
##  Pearson's product-moment correlation
## 
## data:  data_clean$SPV_individual and data_clean$SPV_meta_R
## t = 13.19, df = 776, p-value < 2.2e-16
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
##  0.3687373 0.4836738
## sample estimates:
##       cor 
## 0.4279341

# B: SPV v. ft_diff_out
cor.test(data_clean$SPV_individual, data_clean$ft_diff_out)

## 
##  Pearson's product-moment correlation
## 
## data:  data_clean$SPV_individual and data_clean$ft_diff_out
## t = 1.4789, df = 773, p-value = 0.1396
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
##  -0.01737027  0.12308203
## sample estimates:
##        cor 
## 0.05311858

# C: SPV v. ft_diff_in
cor.test(data_clean$SPV_individual, data_clean$ft_diff_in)

## 
##  Pearson's product-moment correlation
## 
## data:  data_clean$SPV_individual and data_clean$ft_diff_in
## t = 1.7228, df = 775, p-value = 0.08533
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
##  -0.00860548  0.13152680
## sample estimates:
##        cor 
## 0.06176504

# D: in-party meta-SPV v. ft_diff_in
cor.test(data_clean$SPV_meta_R, data_clean$ft_diff_in)

## 
##  Pearson's product-moment correlation
## 
## data:  data_clean$SPV_meta_R and data_clean$ft_diff_in
## t = 1.4783, df = 781, p-value = 0.1397
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
##  -0.01730342  0.12243347
## sample estimates:
##        cor 
## 0.05282361

# E: out-party meta-SPV v. ft_diff_out
cor.test(data_clean$SPV_meta_R, data_clean$ft_diff_out)

## 
##  Pearson's product-moment correlation
## 
## data:  data_clean$SPV_meta_R and data_clean$ft_diff_out
## t = 0.70495, df = 778, p-value = 0.4811
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
##  -0.04501197  0.09529419
## sample estimates:
##        cor 
## 0.02526553

I will also run two correlations designed to test whether the four SPV questions that are more concretely scenario-based correlate to the two SPV questions that are broader and vaguer in their depicted scenarios, allowing us to determine whether there may be any confounding effects of question wording.

cor.test(data_clean$SPV_indiv_scenario, data_clean$SPV_indiv_broad)

## 
##  Pearson's product-moment correlation
## 
## data:  data_clean$SPV_indiv_scenario and data_clean$SPV_indiv_broad
## t = 27.639, df = 777, p-value < 2.2e-16
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
##  0.6668296 0.7378436
## sample estimates:
##       cor 
## 0.7040927

cor.test(data_clean$SPV_meta_scenario, data_clean$SPV_meta_broad)

## 
##  Pearson's product-moment correlation
## 
## data:  data_clean$SPV_meta_scenario and data_clean$SPV_meta_broad
## t = 25.556, df = 779, p-value < 2.2e-16
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
##  0.6352486 0.7117409
## sample estimates:
##       cor 
## 0.6753066

I deviate from the analysis plan by constructing correlation matrices for all pre-treatment and post-treatment variables separately. All correlations reported above are still included.

pretreat_matrix <- data_clean %>%
  dplyr::select (condition_dummy, 
                 party_numeric, ideology_selfreport, maga, PID_strength, social_closeness, ft_rep_pre, ft_dem_pre)
pretreat_results <- round(cor(pretreat_matrix, use = "pairwise.complete.obs"), 2)
pretreat_stars <- cor_with_stars(pretreat_matrix)

	condition_dummy	party_numeric	ideology_selfreport	maga	PID_strength	social_closeness	ft_rep_pre	ft_dem_pre
condition_dummy	—	-0.07*	-0.11**	-0.08*	0.00	-0.01	-0.10**	0.05
party_numeric	-0.07*	—	0.85***	0.53***	-0.06	0.33***	0.79***	-0.65***
ideology_selfreport	-0.11**	0.85***	—	0.50***	0.01	0.36***	0.77***	-0.56***
maga	-0.08*	0.53***	0.50***	—	0.14***	0.10**	0.47***	-0.46***
PID_strength	0.00	-0.06	0.01	0.14***	—	-0.16***	0.10**	0.29***
social_closeness	-0.01	0.33***	0.36***	0.10**	-0.16***	—	0.51***	-0.05
ft_rep_pre	-0.10**	0.79***	0.77***	0.47***	0.10**	0.51***	—	-0.42***
ft_dem_pre	0.05	-0.65***	-0.56***	-0.46***	0.29***	-0.05	-0.42***	—

postreat_matrix <- data_clean %>%
  dplyr::select (condition_dummy,
                 ft_rep_post, ft_dem_post, ft_diff_in, ft_diff_out,
                 moral_disengagement_outgroup, SPV_individual, SPV_meta_R, SPV_indiv_scenario, SPV_indiv_broad, SPV_meta_scenario, SPV_meta_broad)
postreat_results <- round(cor(postreat_matrix, use = "pairwise.complete.obs"), 2)
postreat_stars <- cor_with_stars(postreat_matrix)

	condition_dummy	ft_rep_post	ft_dem_post	ft_diff_in	ft_diff_out	moral_disengagement_outgroup	SPV_individual	SPV_meta_R	SPV_indiv_scenario	SPV_indiv_broad	SPV_meta_scenario	SPV_meta_broad
condition_dummy	—	-0.13***	0.05	-0.01	-0.12**	0.05	0.06	0.11**	0.06	0.04	0.11**	0.09*
ft_rep_post	-0.13***	—	-0.44***	-0.07	0.07	-0.36***	-0.12***	-0.31***	-0.11**	-0.11**	-0.21***	-0.41***
ft_dem_post	0.05	-0.44***	—	0.16***	-0.04	-0.01	0.01	0.12***	0.02	0.01	0.06	0.20***
ft_diff_in	-0.01	-0.07	0.16***	—	-0.00	0.11**	0.06	0.05	0.06	0.04	0.04	0.06
ft_diff_out	-0.12**	0.07	-0.04	-0.00	—	-0.05	0.05	0.03	0.06	0.04	0.04	0.02
moral_disengagement_outgroup	0.05	-0.36***	-0.01	0.11**	-0.05	—	0.30***	0.24***	0.29***	0.27***	0.18***	0.27***
SPV_individual	0.06	-0.12***	0.01	0.06	0.05	0.30***	—	0.43***	0.96***	0.87***	0.41***	0.36***
SPV_meta_R	0.11**	-0.31***	0.12***	0.05	0.03	0.24***	0.43***	—	0.41***	0.38***	0.95***	0.86***
SPV_indiv_scenario	0.06	-0.11**	0.02	0.06	0.06	0.29***	0.96***	0.41***	—	0.70***	0.41***	0.31***
SPV_indiv_broad	0.04	-0.11**	0.01	0.04	0.04	0.27***	0.87***	0.38***	0.70***	—	0.32***	0.40***
SPV_meta_scenario	0.11**	-0.21***	0.06	0.04	0.04	0.18***	0.41***	0.95***	0.41***	0.32***	—	0.68***
SPV_meta_broad	0.09*	-0.41***	0.20***	0.06	0.02	0.27***	0.36***	0.86***	0.31***	0.40***	0.68***	—

Furthermore, I will also run a correlation analysis between each of the eight emotions named in the PANAS-M—Anger, Shame, Guilt, Disgust, Fear, Anxiety, Sadness, and Happiness—and our pre-treatment measures of partisan self-concept, social closeness, affective ratings toward the Republican Party, and affective ratings toward the Democratic Party. I selected only pre-treatment measures since participants will only complete the PANAS-M scale following random assignment to the treatment or control group.

Here, I diverge from the pre-analysis plan to prevent any effects of post-treatment conditioning. That is, I only report correlations between the treatment and post-treatment variables such that the pre-treatment variables are not interpreted as corollaries for the significance of a treatment effect on any post-treatment variables.

emotion_matrix <- data_clean %>%
  dplyr::select (condition_dummy,
                 angry_dummy,
                 shame_dummy,
                 guilt_dummy,
                 disgust_dummy,
                 fear_dummy,
                 anxiety_dummy,
                 sadness_dummy,
                 happiness_dummy)
emotion_results <- round(cor(emotion_matrix, use = "pairwise.complete.obs"), 2)
emotion_stars <- cor_with_stars(emotion_matrix)

	condition_dummy	angry_dummy	shame_dummy	guilt_dummy	disgust_dummy	fear_dummy	anxiety_dummy	sadness_dummy	happiness_dummy
condition_dummy	—	0.52***	0.16***	-0.04	0.63***	0.19***	0.02	0.21***	-0.36***
angry_dummy	0.52***	—	0.20***	0.05	0.68***	0.36***	0.25***	0.29***	-0.32***
shame_dummy	0.16***	0.20***	—	0.33***	0.25***	0.23***	0.16***	0.29***	-0.17***
guilt_dummy	-0.04	0.05	0.33***	—	0.06	0.22***	0.22***	0.18***	-0.11**
disgust_dummy	0.63***	0.68***	0.25***	0.06	—	0.32***	0.18***	0.35***	-0.42***
fear_dummy	0.19***	0.36***	0.23***	0.22***	0.32***	—	0.44***	0.35***	-0.20***
anxiety_dummy	0.02	0.25***	0.16***	0.22***	0.18***	0.44***	—	0.31***	-0.24***
sadness_dummy	0.21***	0.29***	0.29***	0.18***	0.35***	0.35***	0.31***	—	-0.28***
happiness_dummy	-0.36***	-0.32***	-0.17***	-0.11**	-0.42***	-0.20***	-0.24***	-0.28***	—

Conditional Average Treatment Effect (CATE)

Broadly speaking, all analyses will test the effect of shared partisan identity across the pre-identified set of hypothesis-specific outcome variables through a set of Ordinary Least Squares (OLS) linear regression models with heteroskedasticity-robust standard errors applied. This method will determine whether the contrast between Democratic and Republican participants varies in overall magnitude between the treatment and control groups.

library(readxl)
library(tidyverse)
library(psych)
library(interactions)
library(broom)
library(emmeans)
library(stargazer)
library(effsize)
library(modelsummary)
library(lsr)
library(rempsyc)
library(estimatr)
library(ggsignif)

To test Hypotheses 1a through 3b, I will conduct an Ordinary Least Squares (OLS) linear regression model with heteroskedasticity-robust standard errors applied. For Hypotheses 1a and 1b, I will regress the composite variable for meta-perceptions of the average Republican’s support for political violence onto a dichotomous variable denoting condition assignment, with the numeric version of the ANES 7-point party identification scale and the interaction between condition assignment and the numeric version of the ANES 7-point party identification scale as the two covariate terms.

H1<- lm_robust(SPV_meta_R ~ condition_dummy * party_numeric,
                       data = data_clean, se_type = "HC2")

term	estimate	std.error	statistic	p.value	conf.low	conf.high	df	outcome
(Intercept)	45.096	2.185	20.639	0.000	40.807	49.386	781	SPV_meta_R
condition_dummy	6.309	3.255	1.938	0.053	-0.080	12.698	781	SPV_meta_R
party_numeric	-2.298	0.482	-4.769	0.000	-3.244	-1.352	781	SPV_meta_R
condition_dummy:party_numeric	-0.480	0.774	-0.621	0.535	-2.000	1.039	781	SPV_meta_R

To test Hypotheses 2a and 2b, I will regress the composite variable for moral disengagement toward the outgroup party onto a dichotomous variable denoting condition assignment, with the numeric version of the ANES 7-point party identification scale and the interaction between condition assignment and the numeric version of the ANES 7-point party identification scale as the two covariate terms.

H2<- lm_robust(moral_disengagement_outgroup ~ condition_dummy * party_numeric,
                       data = data_clean, se_type = "HC2")

term	estimate	std.error	statistic	p.value	conf.low	conf.high	df	outcome
(Intercept)	3.521	0.088	40.227	0.000	3.349	3.693	785	moral_disengagement_outgroup
condition_dummy	0.108	0.129	0.838	0.402	-0.145	0.360	785	moral_disengagement_outgroup
party_numeric	-0.113	0.025	-4.520	0.000	-0.162	-0.064	785	moral_disengagement_outgroup
condition_dummy:party_numeric	-0.008	0.037	-0.217	0.828	-0.081	0.065	785	moral_disengagement_outgroup

To test Hypotheses 3a and 3b, I will regress the composite variable for individual support for political violence toward the outgroup party onto a dichotomous variable denoting condition assignment, with the numeric version of the ANES 7-point party identification scale and the interaction between condition assignment and the numeric version of the ANES 7-point party identification scale as the two covariate terms.

H3<- lm_robust(SPV_individual ~ condition_dummy * party_numeric,
                     data = data_clean, se_type = "HC2")

term	estimate	std.error	statistic	p.value	conf.low	conf.high	df	outcome
(Intercept)	22.109	1.660	13.319	0.000	18.850	25.367	775	SPV_individual
condition_dummy	4.957	2.470	2.006	0.045	0.107	9.806	775	SPV_individual
party_numeric	-0.216	0.455	-0.474	0.636	-1.109	0.678	775	SPV_individual
condition_dummy:party_numeric	-0.912	0.684	-1.333	0.183	-2.256	0.431	775	SPV_individual

Secondary hypothesis tests will examine the significance of the treatment effect on post-treatment affective ratings by an Ordinary Least Squares (OLS) linear regression model with heteroskedasticity-robust standard errors applied. I will test whether there is an overall decrease in affective ratings toward the Republican Party among participants presented with the violence-endorsing message (H4a) by regressing post-treatment ratings toward the Republican Party onto a dichotomous variable representing the two experimental conditions, interacted with the numeric version of the ANES 7-point party identification scale.

H4a<- lm_robust(ft_rep_post ~ condition_dummy * party_numeric,
                        data = data_clean, se_type = "HC2")

term	estimate	std.error	statistic	p.value	conf.low	conf.high	df	outcome
(Intercept)	2.740	1.763	1.554	0.121	-0.721	6.200	783	ft_rep_post
condition_dummy	-4.922	2.369	-2.078	0.038	-9.573	-0.272	783	ft_rep_post
party_numeric	11.427	0.387	29.534	0.000	10.668	12.187	783	ft_rep_post
condition_dummy:party_numeric	0.118	0.556	0.212	0.832	-0.973	1.209	783	ft_rep_post

For Hypothesis 4b, I add the interaction between the dichotomous treatment variable and pre-treatment ratings toward the Republican Party as an additional covariate of interest.

H4b <- lm_robust(ft_rep_post ~ condition_dummy * party_numeric + condition_dummy * ft_rep_pre,
                                 data = data_clean, se_type = "HC2")

term	estimate	std.error	statistic	p.value	conf.low	conf.high	df	outcome
(Intercept)	-0.249	0.447	-0.557	0.578	-1.127	0.629	780	ft_rep_post
condition_dummy	-2.175	0.673	-3.230	0.001	-3.497	-0.853	780	ft_rep_post
party_numeric	0.831	0.315	2.639	0.008	0.213	1.449	780	ft_rep_post
ft_rep_pre	0.934	0.023	41.335	0.000	0.889	0.978	780	ft_rep_post
condition_dummy:party_numeric	0.308	0.449	0.687	0.492	-0.572	1.189	780	ft_rep_post
condition_dummy:ft_rep_pre	-0.023	0.033	-0.703	0.483	-0.088	0.042	780	ft_rep_post

For Hypotheses 5a through 5c, I will conduct a baseline OLS linear regression model with heteroskedasticity-robust standard errors, regressing post-treatment affective ratings toward the Republican Party onto a three-way interaction between the dichotomous treatment variable, the numeric version of the ANES 7-point party identification scale, and a numeric variable representing the three groups of partisan self-concept.

H5 <- lm_robust(ft_rep_post ~ condition_dummy * party_numeric * PID_bin,
                            data = data_clean, se_type = "HC2")

term	estimate	std.error	statistic	p.value	conf.low	conf.high	df	outcome
(Intercept)	-2.831	5.429	-0.522	0.602	-13.488	7.825	775	ft_rep_post
condition_dummy	-10.179	7.768	-1.310	0.190	-25.428	5.071	775	ft_rep_post
party_numeric	9.448	1.478	6.392	0.000	6.547	12.349	775	ft_rep_post
PID_bin	4.044	3.119	1.296	0.195	-2.079	10.167	775	ft_rep_post
condition_dummy:party_numeric	1.048	2.063	0.508	0.611	-3.002	5.099	775	ft_rep_post
condition_dummy:PID_bin	2.699	4.391	0.615	0.539	-5.920	11.318	775	ft_rep_post
party_numeric:PID_bin	1.238	0.812	1.524	0.128	-0.357	2.833	775	ft_rep_post
condition_dummy:party_numeric:PID_bin	-0.345	1.142	-0.302	0.763	-2.588	1.897	775	ft_rep_post

For Hypotheses 6a through 6c, I will regress post-treatment affective ratings toward the Democratic Party onto the same variables identified for testing Hypotheses 5a through 5c.

H6 <- lm_robust(ft_dem_post ~ condition_dummy * party_numeric * PID_bin,
                            data = data_clean, se_type = "HC2")

term	estimate	std.error	statistic	p.value	conf.low	conf.high	df	outcome
(Intercept)	47.634	7.376	6.458	0.000	33.154	62.113	775	ft_dem_post
condition_dummy	-1.507	9.832	-0.153	0.878	-20.808	17.794	775	ft_dem_post
party_numeric	-2.290	2.191	-1.045	0.296	-6.591	2.011	775	ft_dem_post
PID_bin	22.525	3.958	5.691	0.000	14.755	30.295	775	ft_dem_post
condition_dummy:party_numeric	1.422	2.855	0.498	0.619	-4.184	7.027	775	ft_dem_post
condition_dummy:PID_bin	1.531	5.283	0.290	0.772	-8.839	11.901	775	ft_dem_post
party_numeric:PID_bin	-3.236	1.187	-2.727	0.007	-5.565	-0.906	775	ft_dem_post
condition_dummy:party_numeric:PID_bin	-1.151	1.574	-0.731	0.465	-4.239	1.938	775	ft_dem_post

Deviations from PAP

I convert party_numeric into a categorical variable to determine whether the above model-based predictions correspond to the raw data.

Hypothesis 1

H1_factor <- lm_robust(SPV_meta_R ~ condition_dummy * factor(party_numeric),
                       data = data_clean, se_type = "HC2")

term	estimate	std.error	statistic	p.value	conf.low	conf.high	df	outcome
(Intercept)	42.221	2.174	19.420	0.000	37.953	46.489	773	SPV_meta_R
condition_dummy	2.437	3.244	0.751	0.453	-3.930	8.805	773	SPV_meta_R
factor(party_numeric)2	-1.669	3.714	-0.449	0.653	-8.959	5.621	773	SPV_meta_R
factor(party_numeric)3	0.329	4.590	0.072	0.943	-8.681	9.339	773	SPV_meta_R
factor(party_numeric)5	-11.253	3.849	-2.924	0.004	-18.809	-3.698	773	SPV_meta_R
factor(party_numeric)6	-15.086	3.209	-4.701	0.000	-21.386	-8.787	773	SPV_meta_R
factor(party_numeric)7	-10.043	3.529	-2.846	0.005	-16.969	-3.116	773	SPV_meta_R
condition_dummy:factor(party_numeric)2	7.684	5.320	1.444	0.149	-2.759	18.126	773	SPV_meta_R
condition_dummy:factor(party_numeric)3	4.963	6.450	0.770	0.442	-7.698	17.625	773	SPV_meta_R
condition_dummy:factor(party_numeric)5	4.695	7.355	0.638	0.523	-9.743	19.133	773	SPV_meta_R
condition_dummy:factor(party_numeric)6	0.599	4.988	0.120	0.904	-9.193	10.391	773	SPV_meta_R
condition_dummy:factor(party_numeric)7	-2.788	5.945	-0.469	0.639	-14.458	8.882	773	SPV_meta_R

emm_H1_factor <- emmeans(H1_factor, ~ condition_dummy | party_numeric,
                         vcov. = vcov(H1_factor))
contr_H1_factor <- contrast(emm_H1_factor,
                            method = list("Treatment - Control" = c(-1, 1)),
                            by = "party_numeric")
contr_H1_factor_df <- as.data.frame(confint(contr_H1_factor))

contrast	party_numeric	estimate	SE	df	t.ratio	p.value
Treatment - Control	1	2.437	3.244	773	0.751	0.453
Treatment - Control	2	10.121	4.216	773	2.401	0.017
Treatment - Control	3	7.401	5.575	773	1.327	0.185
Treatment - Control	5	7.132	6.601	773	1.081	0.280
Treatment - Control	6	3.037	3.789	773	0.801	0.423
Treatment - Control	7	-0.350	4.982	773	-0.070	0.944

Hypothesis 2

H2_factor <- lm_robust(moral_disengagement_outgroup ~ condition_dummy * factor(party_numeric),
                       data = data_clean, se_type = "HC2")

term	estimate	std.error	statistic	p.value	conf.low	conf.high	df	outcome
(Intercept)	3.530	0.081	43.449	0.000	3.371	3.690	777	moral_disengagement_outgroup
condition_dummy	-0.035	0.120	-0.291	0.771	-0.271	0.201	777	moral_disengagement_outgroup
factor(party_numeric)2	-0.361	0.142	-2.553	0.011	-0.639	-0.084	777	moral_disengagement_outgroup
factor(party_numeric)3	-0.395	0.171	-2.319	0.021	-0.730	-0.061	777	moral_disengagement_outgroup
factor(party_numeric)5	-0.810	0.237	-3.412	0.001	-1.276	-0.344	777	moral_disengagement_outgroup
factor(party_numeric)6	-1.051	0.172	-6.109	0.000	-1.388	-0.713	777	moral_disengagement_outgroup
factor(party_numeric)7	-0.375	0.181	-2.074	0.038	-0.730	-0.020	777	moral_disengagement_outgroup
condition_dummy:factor(party_numeric)2	0.405	0.212	1.916	0.056	-0.010	0.820	777	moral_disengagement_outgroup
condition_dummy:factor(party_numeric)3	0.169	0.233	0.727	0.468	-0.288	0.626	777	moral_disengagement_outgroup
condition_dummy:factor(party_numeric)5	-0.335	0.335	-1.001	0.317	-0.994	0.323	777	moral_disengagement_outgroup
condition_dummy:factor(party_numeric)6	-0.015	0.247	-0.061	0.951	-0.500	0.470	777	moral_disengagement_outgroup
condition_dummy:factor(party_numeric)7	0.316	0.262	1.207	0.228	-0.198	0.830	777	moral_disengagement_outgroup

emm_H2_factor <- emmeans(H2_factor, ~ condition_dummy | party_numeric,
                         vcov. = vcov(H2_factor))
contr_H2_factor <- contrast(emm_H2_factor,
                            method = list("Treatment - Control" = c(-1, 1)),
                            by = "party_numeric")
contr_H2_factor_df <- as.data.frame(confint(contr_H2_factor))

contrast	party_numeric	estimate	SE	df	t.ratio	p.value
Treatment - Control	1	-0.035	0.120	777	-0.291	0.771
Treatment - Control	2	0.370	0.174	777	2.128	0.034
Treatment - Control	3	0.134	0.199	777	0.673	0.501
Treatment - Control	5	-0.370	0.313	777	-1.184	0.237
Treatment - Control	6	-0.050	0.216	777	-0.232	0.817
Treatment - Control	7	0.281	0.233	777	1.209	0.227

Hypothesis 3

H3_factor <- lm_robust(SPV_individual ~ condition_dummy * factor(party_numeric),
                       data = data_clean, se_type = "HC2")

term	estimate	std.error	statistic	p.value	conf.low	conf.high	df	outcome
(Intercept)	22.551	1.661	13.576	0.000	19.290	25.811	767	SPV_individual
condition_dummy	1.466	2.405	0.610	0.542	-3.255	6.187	767	SPV_individual
factor(party_numeric)2	-1.403	2.461	-0.570	0.569	-6.234	3.427	767	SPV_individual
factor(party_numeric)3	-1.114	3.464	-0.322	0.748	-7.915	5.686	767	SPV_individual
factor(party_numeric)5	-3.185	4.237	-0.752	0.452	-11.503	5.133	767	SPV_individual
factor(party_numeric)6	-4.958	2.867	-1.729	0.084	-10.586	0.670	767	SPV_individual
factor(party_numeric)7	1.479	3.624	0.408	0.683	-5.635	8.594	767	SPV_individual
condition_dummy:factor(party_numeric)2	6.180	3.840	1.609	0.108	-1.358	13.718	767	SPV_individual
condition_dummy:factor(party_numeric)3	2.946	4.741	0.621	0.535	-6.361	12.253	767	SPV_individual
condition_dummy:factor(party_numeric)5	-4.624	6.547	-0.706	0.480	-17.476	8.229	767	SPV_individual
condition_dummy:factor(party_numeric)6	-4.370	4.044	-1.080	0.280	-12.309	3.570	767	SPV_individual
condition_dummy:factor(party_numeric)7	-1.111	5.637	-0.197	0.844	-12.178	9.955	767	SPV_individual

emm_H3_factor <- emmeans(H3_factor, ~ condition_dummy | party_numeric,
                         vcov. = vcov(H3_factor))
contr_H3_factor <- contrast(emm_H3_factor,
                            method = list("Treatment - Control" = c(-1, 1)),
                            by = "party_numeric")
contr_H3_factor_df <- as.data.frame(confint(contr_H3_factor))

contrast	party_numeric	estimate	SE	df	t.ratio	p.value
Treatment - Control	1	1.466	2.405	767	0.610	0.542
Treatment - Control	2	7.647	2.994	767	2.554	0.011
Treatment - Control	3	4.412	4.086	767	1.080	0.281
Treatment - Control	5	-3.157	6.090	767	-0.518	0.604
Treatment - Control	6	-2.903	3.252	767	-0.893	0.372
Treatment - Control	7	0.355	5.099	767	0.070	0.944

Hypothesis 4a

H4a_factor <- lm_robust(ft_rep_post ~ condition_dummy * factor(party_numeric),
                        data = data_clean, se_type = "HC2")

term	estimate	std.error	statistic	p.value	conf.low	conf.high	df	outcome
(Intercept)	17.937	1.724	10.406	0.000	14.553	21.321	775	ft_rep_post
condition_dummy	-5.210	2.206	-2.361	0.018	-9.542	-0.879	775	ft_rep_post
factor(party_numeric)2	6.907	2.923	2.363	0.018	1.168	12.646	775	ft_rep_post
factor(party_numeric)3	4.616	3.592	1.285	0.199	-2.435	11.667	775	ft_rep_post
factor(party_numeric)5	42.676	3.853	11.077	0.000	35.113	50.239	775	ft_rep_post
factor(party_numeric)6	54.958	2.985	18.409	0.000	49.097	60.818	775	ft_rep_post
factor(party_numeric)7	67.304	2.550	26.389	0.000	62.298	72.311	775	ft_rep_post
condition_dummy:factor(party_numeric)2	2.687	4.092	0.657	0.512	-5.346	10.721	775	ft_rep_post
condition_dummy:factor(party_numeric)3	1.376	4.693	0.293	0.769	-7.836	10.589	775	ft_rep_post
condition_dummy:factor(party_numeric)5	-3.753	6.197	-0.606	0.545	-15.917	8.412	775	ft_rep_post
condition_dummy:factor(party_numeric)6	-2.371	4.050	-0.585	0.558	-10.321	5.579	775	ft_rep_post
condition_dummy:factor(party_numeric)7	7.398	3.627	2.039	0.042	0.277	14.518	775	ft_rep_post

emm_H4a_factor <- emmeans(H4a_factor, ~ condition_dummy | party_numeric,
                         vcov. = vcov(H4a_factor))
contr_H4a_factor <- contrast(emm_H4a_factor,
                            method = list("Treatment - Control" = c(-1, 1)),
                            by = "party_numeric")
contr_H4a_factor_df <- as.data.frame(confint(contr_H4a_factor))

contrast	party_numeric	estimate	SE	df	t.ratio	p.value
Treatment - Control	1	-5.210	2.206	775	-2.361	0.018
Treatment - Control	2	-2.523	3.447	775	-0.732	0.464
Treatment - Control	3	-3.834	4.142	775	-0.926	0.355
Treatment - Control	5	-8.963	5.791	775	-1.548	0.122
Treatment - Control	6	-7.581	3.396	775	-2.232	0.026
Treatment - Control	7	2.187	2.879	775	0.760	0.448

Hypothesis 4b

H4b_factor <- lm_robust(ft_rep_post ~ condition_dummy * factor(party_numeric)
                                      + condition_dummy * ft_rep_pre,
                                      data = data_clean, se_type = "HC2")

term	estimate	std.error	statistic	p.value	conf.low	conf.high	df	outcome
(Intercept)	0.786	0.379	2.074	0.038	0.042	1.529	772	ft_rep_post
condition_dummy	-1.210	0.601	-2.013	0.044	-2.390	-0.030	772	ft_rep_post
factor(party_numeric)2	-0.259	0.820	-0.315	0.753	-1.869	1.352	772	ft_rep_post
factor(party_numeric)3	2.267	1.341	1.691	0.091	-0.365	4.899	772	ft_rep_post
factor(party_numeric)5	0.970	1.767	0.549	0.583	-2.499	4.439	772	ft_rep_post
factor(party_numeric)6	4.057	1.724	2.353	0.019	0.672	7.442	772	ft_rep_post
factor(party_numeric)7	4.638	1.983	2.339	0.020	0.745	8.530	772	ft_rep_post
ft_rep_pre	0.938	0.022	41.802	0.000	0.894	0.982	772	ft_rep_post
condition_dummy:factor(party_numeric)2	0.911	1.216	0.749	0.454	-1.477	3.298	772	ft_rep_post
condition_dummy:factor(party_numeric)3	-2.036	1.611	-1.264	0.207	-5.199	1.127	772	ft_rep_post
condition_dummy:factor(party_numeric)5	0.645	3.440	0.187	0.851	-6.109	7.398	772	ft_rep_post
condition_dummy:factor(party_numeric)6	1.265	2.557	0.495	0.621	-3.754	6.285	772	ft_rep_post
condition_dummy:factor(party_numeric)7	5.139	3.099	1.658	0.098	-0.944	11.222	772	ft_rep_post
condition_dummy:ft_rep_pre	-0.044	0.035	-1.251	0.211	-0.114	0.025	772	ft_rep_post

emm_H4b_factor <- emmeans(H4b_factor, ~ condition_dummy | party_numeric,
                         vcov. = vcov(H4b_factor))
contr_H4b_factor <- contrast(emm_H4b_factor,
                            method = list("Treatment - Control" = c(-1, 1)),
                            by = "party_numeric")
contr_H4b_factor_df <- as.data.frame(confint(contr_H4b_factor))

contrast	party_numeric	estimate	SE	df	t.ratio	p.value
Treatment - Control	1	-2.883	1.170	772	-2.464	0.014
Treatment - Control	2	-1.973	1.179	772	-1.673	0.095
Treatment - Control	3	-4.920	1.619	772	-3.039	0.002
Treatment - Control	5	-2.239	3.221	772	-0.695	0.487
Treatment - Control	6	-1.618	1.721	772	-0.940	0.347
Treatment - Control	7	2.256	2.133	772	1.058	0.291

For the following hypothesis groups, I collapsed Weak and Medium Partisan Self-Concept into the same group given the low N.

data_clean <- data_clean |>
  mutate(
    PID_bin_collapsed = case_when(
      PID_bin %in% c(0, 1) ~ "Weak-Medium",
      PID_bin == 2          ~ "Strong",
      TRUE                  ~ NA_character_
    ),
    PID_bin_collapsed = factor(PID_bin_collapsed,
                               levels = c("Weak-Medium", "Strong"))
  )

Hypothesis 5

H5_factor <- lm_robust(ft_rep_post ~ condition_dummy * factor(party_numeric) * PID_bin_collapsed,
                       data = data_clean, se_type = "HC2")

term	estimate	std.error	statistic	p.value	conf.low	conf.high	df	outcome
(Intercept)	12.474	2.640	4.726	0.000	7.292	17.655	759	ft_rep_post
condition_dummy	-4.768	3.731	-1.278	0.202	-12.092	2.557	759	ft_rep_post
factor(party_numeric)2	12.145	4.114	2.952	0.003	4.070	20.221	759	ft_rep_post
factor(party_numeric)3	10.453	4.280	2.442	0.015	2.050	18.856	759	ft_rep_post
factor(party_numeric)5	46.633	4.529	10.296	0.000	37.742	55.525	759	ft_rep_post
factor(party_numeric)6	52.860	4.936	10.709	0.000	43.169	62.550	759	ft_rep_post
factor(party_numeric)7	59.360	7.852	7.559	0.000	43.944	74.775	759	ft_rep_post
PID_bin_collapsedStrong	6.424	3.289	1.953	0.051	-0.032	12.881	759	ft_rep_post
condition_dummy:factor(party_numeric)2	-0.276	5.863	-0.047	0.962	-11.786	11.233	759	ft_rep_post
condition_dummy:factor(party_numeric)3	0.433	5.832	0.074	0.941	-11.017	11.882	759	ft_rep_post
condition_dummy:factor(party_numeric)5	-4.692	6.739	-0.696	0.486	-17.922	8.537	759	ft_rep_post
condition_dummy:factor(party_numeric)6	3.353	6.143	0.546	0.585	-8.705	15.412	759	ft_rep_post
condition_dummy:factor(party_numeric)7	-20.066	9.676	-2.074	0.038	-39.059	-1.072	759	ft_rep_post
condition_dummy:PID_bin_collapsedStrong	-0.480	4.486	-0.107	0.915	-9.288	8.327	759	ft_rep_post
factor(party_numeric)2:PID_bin_collapsedStrong	-5.985	5.879	-1.018	0.309	-17.526	5.557	759	ft_rep_post
factor(party_numeric)3:PID_bin_collapsedStrong	-9.351	10.747	-0.870	0.385	-30.449	11.747	759	ft_rep_post
factor(party_numeric)5:PID_bin_collapsedStrong	9.135	6.962	1.312	0.190	-4.533	22.803	759	ft_rep_post
factor(party_numeric)6:PID_bin_collapsedStrong	7.942	5.710	1.391	0.165	-3.267	19.151	759	ft_rep_post
factor(party_numeric)7:PID_bin_collapsedStrong	8.781	8.301	1.058	0.290	-7.515	25.078	759	ft_rep_post
condition_dummy:factor(party_numeric)2:PID_bin_collapsedStrong	5.466	8.302	0.658	0.511	-10.833	21.764	759	ft_rep_post
condition_dummy:factor(party_numeric)3:PID_bin_collapsedStrong	4.315	13.350	0.323	0.747	-21.893	30.524	759	ft_rep_post
condition_dummy:factor(party_numeric)5:PID_bin_collapsedStrong	-1.726	23.623	-0.073	0.942	-48.101	44.649	759	ft_rep_post
condition_dummy:factor(party_numeric)6:PID_bin_collapsedStrong	-8.805	8.850	-0.995	0.320	-26.178	8.567	759	ft_rep_post
condition_dummy:factor(party_numeric)7:PID_bin_collapsedStrong	27.724	10.307	2.690	0.007	7.490	47.959	759	ft_rep_post

emm_H5_factor <- emmeans(H5_factor, ~ condition_dummy | party_numeric + PID_bin_collapsed,
                         vcov. = vcov(H5_factor))
contr_H5_factor <- contrast(emm_H5_factor,
                            method = list("Treatment - Control" = c(-1, 1)),
                            by = c("party_numeric", "PID_bin_collapsed"))
contr_H5_factor_df <- as.data.frame(confint(contr_H5_factor))

contrast	party_numeric	PID_bin_collapsed	estimate	SE	df	t.ratio	p.value
Treatment - Control	1	Weak-Medium	-4.768	3.731	759	-1.278	0.202
Treatment - Control	2	Weak-Medium	-5.044	4.522	759	-1.115	0.265
Treatment - Control	3	Weak-Medium	-4.335	4.483	759	-0.967	0.334
Treatment - Control	5	Weak-Medium	-9.460	5.612	759	-1.686	0.092
Treatment - Control	6	Weak-Medium	-1.414	4.880	759	-0.290	0.772
Treatment - Control	7	Weak-Medium	-24.833	8.927	759	-2.782	0.006
Treatment - Control	1	Strong	-5.248	2.491	759	-2.107	0.035
Treatment - Control	2	Strong	-0.059	5.325	759	-0.011	0.991
Treatment - Control	3	Strong	-0.500	11.748	759	-0.043	0.966
Treatment - Control	5	Strong	-11.667	22.504	759	-0.518	0.604
Treatment - Control	6	Strong	-10.700	5.863	759	-1.825	0.068
Treatment - Control	7	Strong	2.411	2.533	759	0.952	0.342

Hypothesis 6

H6_factor <- lm_robust(ft_dem_post ~ condition_dummy * factor(party_numeric) * PID_bin_collapsed,
                       data = data_clean, se_type = "HC2")

term	estimate	std.error	statistic	p.value	conf.low	conf.high	df	outcome
(Intercept)	59.368	4.753	12.490	0.000	50.038	68.699	759	ft_dem_post
condition_dummy	7.749	6.818	1.137	0.256	-5.636	21.134	759	ft_dem_post
factor(party_numeric)2	-2.392	5.562	-0.430	0.667	-13.311	8.527	759	ft_dem_post
factor(party_numeric)3	-7.320	5.773	-1.268	0.205	-18.653	4.013	759	ft_dem_post
factor(party_numeric)5	-29.702	6.808	-4.362	0.000	-43.067	-16.336	759	ft_dem_post
factor(party_numeric)6	-13.405	7.351	-1.824	0.069	-27.835	1.025	759	ft_dem_post
factor(party_numeric)7	-29.535	14.188	-2.082	0.038	-57.387	-1.684	759	ft_dem_post
PID_bin_collapsedStrong	24.280	4.940	4.915	0.000	14.582	33.978	759	ft_dem_post
condition_dummy:factor(party_numeric)2	-7.575	8.073	-0.938	0.348	-23.423	8.273	759	ft_dem_post
condition_dummy:factor(party_numeric)3	-8.084	8.247	-0.980	0.327	-24.273	8.105	759	ft_dem_post
condition_dummy:factor(party_numeric)5	2.584	10.431	0.248	0.804	-17.893	23.061	759	ft_dem_post
condition_dummy:factor(party_numeric)6	-12.423	9.633	-1.290	0.198	-31.334	6.488	759	ft_dem_post
condition_dummy:factor(party_numeric)7	-7.083	18.906	-0.375	0.708	-44.197	30.031	759	ft_dem_post
condition_dummy:PID_bin_collapsedStrong	-7.480	7.105	-1.053	0.293	-21.428	6.468	759	ft_dem_post
factor(party_numeric)2:PID_bin_collapsedStrong	-4.697	6.258	-0.751	0.453	-16.982	7.587	759	ft_dem_post
factor(party_numeric)3:PID_bin_collapsedStrong	-6.162	9.857	-0.625	0.532	-25.513	13.189	759	ft_dem_post
factor(party_numeric)5:PID_bin_collapsedStrong	-5.280	14.478	-0.365	0.715	-33.701	23.142	759	ft_dem_post
factor(party_numeric)6:PID_bin_collapsedStrong	-27.243	8.879	-3.068	0.002	-44.674	-9.811	759	ft_dem_post
factor(party_numeric)7:PID_bin_collapsedStrong	-25.250	14.634	-1.725	0.085	-53.978	3.478	759	ft_dem_post
condition_dummy:factor(party_numeric)2:PID_bin_collapsedStrong	5.260	8.888	0.592	0.554	-12.188	22.707	759	ft_dem_post
condition_dummy:factor(party_numeric)3:PID_bin_collapsedStrong	14.148	12.505	1.131	0.258	-10.401	38.697	759	ft_dem_post
condition_dummy:factor(party_numeric)5:PID_bin_collapsedStrong	-0.187	29.485	-0.006	0.995	-58.068	57.695	759	ft_dem_post
condition_dummy:factor(party_numeric)6:PID_bin_collapsedStrong	9.439	12.872	0.733	0.464	-15.830	34.709	759	ft_dem_post
condition_dummy:factor(party_numeric)7:PID_bin_collapsedStrong	1.551	19.647	0.079	0.937	-37.018	40.119	759	ft_dem_post

emm_H6_factor <- emmeans(H6_factor, ~ condition_dummy | party_numeric + PID_bin_collapsed,
                         vcov. = vcov(H6_factor))
contr_H6_factor <- contrast(emm_H6_factor,
                            method = list("Treatment - Control" = c(-1, 1)),
                            by = c("party_numeric", "PID_bin_collapsed"))
contr_H6_factor_df <- as.data.frame(confint(contr_H6_factor))

contrast	party_numeric	PID_bin_collapsed	estimate	SE	df	t.ratio	p.value
Treatment - Control	1	Weak-Medium	7.749	6.818	759	1.137	0.256
Treatment - Control	2	Weak-Medium	0.174	4.323	759	0.040	0.968
Treatment - Control	3	Weak-Medium	-0.334	4.639	759	-0.072	0.943
Treatment - Control	5	Weak-Medium	10.333	7.894	759	1.309	0.191
Treatment - Control	6	Weak-Medium	-4.673	6.805	759	-0.687	0.492
Treatment - Control	7	Weak-Medium	0.667	17.634	759	0.038	0.970
Treatment - Control	1	Strong	0.269	1.998	759	0.135	0.893
Treatment - Control	2	Strong	-2.047	3.135	759	-0.653	0.514
Treatment - Control	3	Strong	6.333	9.186	759	0.689	0.491
Treatment - Control	5	Strong	2.667	27.506	759	0.097	0.923
Treatment - Control	6	Strong	-2.714	8.301	759	-0.327	0.744
Treatment - Control	7	Strong	-5.263	4.957	759	-1.062	0.289

Planned Contrasts

Across all hypotheses, except Hypothesis 4b (which relies on a multivariate Ordinary Least Squares regression model), I will also employ planned contrasts using estimated marginal means to measure the difference in means between the treatment and control groups. This method will determine the model-based predictions of the average treatment effect for our identified groups of interest, measured as t-ratios and adjusted for covariates and unbalanced sample sizes. For our primary hypotheses (1a through 3b), along with Hypothesis 4a, our planned contrasts are fourfold: Republicans in the treatment condition versus Democrats in the treatment condition, Republicans in the control condition versus Democrats in the control condition, Republicans in the control condition versus Republicans in the treatment condition, and Democrats in the control condition versus Democrats in the treatment condition.

For all hypotheses, I define my groups based on the 7-point party identification scale detailed above, of which the six groups referenced are “Strong D,” “Weak D”, “Lean D”, “Lean R”, “Weak R,” and “Strong R”. As such, I deviate from the pre-analysis plan outlined for Hypotheses 1a through 4b, in which I defined my groups of interest as the control and treatment groups for Republicans and Democrats separately. However, this method allows me to ascertain deeper group-level differences in treatment effects that would otherwise not be observable if comparing treated Democrats to treated Republicans more broadly, for example.

Hypothesis 1

emm_H1<- emmeans(H1, ~ condition_dummy | party_numeric,
                 at = list(party_numeric = c(1, 2, 3, 5, 6, 7)),
                 vcov. = vcov(H1))
contr_H1<- contrast(emm_H1,
                    method = list("Treatment - Control" = c(-1, 1)),
                    by = "party_numeric")
contr_H1_df <- as.data.frame(confint(contr_H1))
contr_H1_table <- summary(contr_H1) |>
  as.data.frame() |>
  dplyr::select(contrast, party_numeric, estimate, SE, df, t.ratio, p.value) |>
  mutate(party_numeric = factor(party_numeric, levels = c(1, 2, 3, 5, 6, 7),
                                labels = c("Strong D","Weak D","Lean D","Lean R","Weak R","Strong R")))

contrast	party_numeric	estimate	SE	df	t.ratio	p.value
Treatment - Control	Strong D	5.828231	2.645362	781	2.2031886	0.0278727
Treatment - Control	Weak D	5.347792	2.145046	781	2.4930894	0.0128697
Treatment - Control	Lean D	4.867352	1.844564	781	2.6387546	0.0084869
Treatment - Control	Lean R	3.906473	2.144420	781	1.8216922	0.0688841
Treatment - Control	Weak R	3.426034	2.644516	781	1.2955241	0.1955223
Treatment - Control	Strong R	2.945595	3.253660	781	0.9053174	0.3655765

pA_H1 <- data_clean |>
  filter(!is.na(exp_group_category)) |>
  group_by(party_numeric, exp_group_category) |>
  summarise(
    mean_y = mean(SPV_meta_R, na.rm = TRUE),
    se_y   = sd(SPV_meta_R, na.rm = TRUE) / sqrt(n()),
    .groups = "drop"
  ) |>
  ggplot(aes(x = factor(party_numeric), y = mean_y, color = exp_group_category)) +
  geom_errorbar(aes(ymin = mean_y - 1.96 * se_y,
                    ymax = mean_y + 1.96 * se_y),
                width = 0.15, linewidth = 0.6,
                position = position_dodge(width = 0.4)) +
  geom_point(size = 3, position = position_dodge(width = 0.4)) +
  scale_x_discrete(labels = pid_labels) +
  scale_color_manual(
    values = c(
      "Control R"   = "darkred",
      "Treatment R" = "hotpink",
      "Control D"   = "midnightblue",
      "Treatment D" = "dodgerblue"
    ),
    name = "Condition"
  ) +
  labs(title = "A", x = NULL,
       y = "Meta-Perceptions of Republican SPV\n(raw mean)") +
  base_theme +
  theme(legend.position = "bottom")

pB_H1 <- ggplot(contr_H1_factor_df, aes(x = party_numeric, y = estimate)) +
  geom_hline(yintercept = 0, linetype = "dashed", color = "grey50") +
  geom_smooth(method = "lm_robust", se = TRUE,
              color = "grey20", fill = "grey60",
              alpha = 0.2, linewidth = 0.9) +
  geom_errorbar(aes(ymin = lower.CL, ymax = upper.CL,
                    color = factor(party_numeric)),
                width = 0.15, linewidth = 0.6) +
  geom_point(aes(color = factor(party_numeric)), size = 3) +
  scale_color_manual(values = pid_colors, guide = "none") +
  scale_x_continuous(breaks = pid_breaks, labels = pid_labels) +
  labs(title = "B", x = "Party ID",
       y = "Difference in Means") +
  base_theme

Hypothesis 2

emm_H2<- emmeans(H2, ~ condition_dummy | party_numeric,
                 at = list(party_numeric = c(1, 2, 3, 5, 6, 7)),
                 vcov. = vcov(H2))
contr_H2<- contrast(emm_H2,
                    method = list("Treatment - Control" = c(-1, 1)),
                    by = "party_numeric")
contr_H2_df <- as.data.frame(confint(contr_H2))
contr_H2_table <- summary(contr_H2) |>
  as.data.frame() |>
  dplyr::select(contrast, party_numeric, estimate, SE, df, t.ratio, p.value) |>
  mutate(party_numeric = factor(party_numeric, levels = c(1, 2, 3, 5, 6, 7),
                                labels = c("Strong D","Weak D","Lean D","Lean R","Weak R","Strong R")))

contrast	party_numeric	estimate	SE	df	t.ratio	p.value
Treatment - Control	Strong D	0.0998034	0.1014437	785	0.9838303	0.3255021
Treatment - Control	Weak D	0.0917773	0.0822541	785	1.1157771	0.2648590
Treatment - Control	Lean D	0.0837512	0.0772990	785	1.0834710	0.2789321
Treatment - Control	Lean R	0.0676990	0.1122377	785	0.6031754	0.5465662
Treatment - Control	Weak R	0.0596729	0.1414570	785	0.4218450	0.6732536
Treatment - Control	Strong R	0.0516469	0.1736587	785	0.2974044	0.7662365

pA_H2 <- data_clean |>
  filter(!is.na(exp_group_category)) |>
  group_by(party_numeric, exp_group_category) |>
  summarise(
    mean_y = mean(moral_disengagement_outgroup, na.rm = TRUE),
    se_y   = sd(moral_disengagement_outgroup, na.rm = TRUE) / sqrt(n()),
    .groups = "drop"
  ) |>
  ggplot(aes(x = factor(party_numeric), y = mean_y, color = exp_group_category)) +
  geom_errorbar(aes(ymin = mean_y - 1.96 * se_y,
                    ymax = mean_y + 1.96 * se_y),
                width = 0.15, linewidth = 0.6,
                position = position_dodge(width = 0.4)) +
  geom_point(size = 3, position = position_dodge(width = 0.4)) +
  scale_x_discrete(labels = pid_labels) +
  scale_color_manual(
    values = c(
      "Control R"   = "darkred",
      "Treatment R" = "hotpink",
      "Control D"   = "midnightblue",
      "Treatment D" = "dodgerblue"
    ),
    name = "Condition"
  ) +
  labs(title = "A", x = NULL,
       y = "Moral Disengagement (Out-Party)\n(raw mean)") +
  base_theme +
  theme(legend.position = "bottom")

pB_H2 <- ggplot(contr_H2_factor_df, aes(x = party_numeric, y = estimate)) +
  geom_hline(yintercept = 0, linetype = "dashed", color = "grey50") +
  geom_smooth(method = "lm_robust", se = TRUE,
              color = "grey20", fill = "grey60",
              alpha = 0.2, linewidth = 0.9) +
  geom_errorbar(aes(ymin = lower.CL, ymax = upper.CL,
                    color = factor(party_numeric)),
                width = 0.15, linewidth = 0.6) +
  geom_point(aes(color = factor(party_numeric)), size = 3) +
  scale_color_manual(values = pid_colors, guide = "none") +
  scale_x_continuous(breaks = pid_breaks, labels = pid_labels) +
  labs(title = "B", x = "Party ID",
       y = "Difference in Means") +
  base_theme

Hypothesis 3

emm_H3<- emmeans(H3, ~ condition_dummy | party_numeric,
                 at = list(party_numeric = c(1, 2, 3, 5, 6, 7)),
                 vcov. = vcov(H3))
contr_H3<- contrast(emm_H3,
                    method = list("Treatment - Control" = c(-1, 1)),
                    by = "party_numeric")
contr_H3_df <- as.data.frame(confint(contr_H3))
contr_H3_table <- summary(contr_H3) |>
  as.data.frame() |>
  dplyr::select(contrast, party_numeric, estimate, SE, df, t.ratio, p.value) |>
  mutate(party_numeric = factor(party_numeric, levels = c(1, 2, 3, 5, 6, 7),
                                labels = c("Strong D","Weak D","Lean D","Lean R","Weak R","Strong R")))

contrast	party_numeric	estimate	SE	df	t.ratio	p.value
Treatment - Control	Strong D	4.0441647	1.954600	775	2.0690494	0.0388725
Treatment - Control	Weak D	3.1317498	1.573260	775	1.9906117	0.0468744
Treatment - Control	Lean D	2.2193348	1.437483	775	1.5439036	0.1230199
Treatment - Control	Lean R	0.3945049	2.014535	775	0.1958293	0.8447951
Treatment - Control	Weak R	-0.5179100	2.541483	775	-0.2037826	0.8385769
Treatment - Control	Strong R	-1.4303249	3.129928	775	-0.4569834	0.6478111

pA_H3 <- data_clean |>
  filter(!is.na(exp_group_category)) |>
  group_by(party_numeric, exp_group_category) |>
  summarise(
    mean_y = mean(SPV_individual, na.rm = TRUE),
    se_y   = sd(SPV_individual, na.rm = TRUE) / sqrt(n()),
    .groups = "drop"
  ) |>
  ggplot(aes(x = factor(party_numeric), y = mean_y, color = exp_group_category)) +
  geom_errorbar(aes(ymin = mean_y - 1.96 * se_y,
                    ymax = mean_y + 1.96 * se_y),
                width = 0.15, linewidth = 0.6,
                position = position_dodge(width = 0.4)) +
  geom_point(size = 3, position = position_dodge(width = 0.4)) +
  scale_x_discrete(labels = pid_labels) +
  scale_color_manual(
    values = c(
      "Control R"   = "darkred",
      "Treatment R" = "hotpink",
      "Control D"   = "midnightblue",
      "Treatment D" = "dodgerblue"
    ),
    name = "Condition"
  ) +
  labs(title = "A", x = NULL,
       y = "Individual SPV\n(raw mean)") +
  base_theme +
  theme(legend.position = "bottom")

pB_H3 <- ggplot(contr_H3_factor_df, aes(x = party_numeric, y = estimate)) +
  geom_hline(yintercept = 0, linetype = "dashed", color = "grey50") +
  geom_smooth(method = "lm_robust", se = TRUE,
              color = "grey20", fill = "grey60",
              alpha = 0.2, linewidth = 0.9) +
  geom_errorbar(aes(ymin = lower.CL, ymax = upper.CL,
                    color = factor(party_numeric)),
                width = 0.15, linewidth = 0.6) +
  geom_point(aes(color = factor(party_numeric)), size = 3) +
  scale_color_manual(values = pid_colors, guide = "none") +
  scale_x_continuous(breaks = pid_breaks, labels = pid_labels) +
  labs(title = "B", x = "Party ID",
       y = "Difference in Means") +
  base_theme

Hypothesis 4a

emm_H4a<- emmeans(H4a, ~ condition_dummy | party_numeric,
                 at = list(party_numeric = c(1, 2, 3, 5, 6, 7)),
                 vcov. = vcov(H4a))
contr_H4a<- contrast(emm_H4a,
                    method = list("Treatment - Control" = c(-1, 1)),
                    by = "party_numeric")
contr_H4a_df <- as.data.frame(confint(contr_H4a))
contr_H4a_table <- summary(contr_H4a) |>
  as.data.frame() |>
  dplyr::select(contrast, party_numeric, estimate, SE, df, t.ratio, p.value) |>
  mutate(party_numeric = factor(party_numeric, levels = c(1, 2, 3, 5, 6, 7),
                                labels = c("Strong D","Weak D","Lean D","Lean R","Weak R","Strong R")))

contrast	party_numeric	estimate	SE	df	t.ratio	p.value
Treatment - Control	Strong D	-4.804423	1.949049	783	-2.465009	0.0139146
Treatment - Control	Weak D	-4.686386	1.613525	783	-2.904440	0.0037827
Treatment - Control	Lean D	-4.568350	1.423416	783	-3.209427	0.0013842
Treatment - Control	Lean R	-4.332277	1.650865	783	-2.624246	0.0088533
Treatment - Control	Weak R	-4.214241	2.000487	783	-2.106607	0.0354692
Treatment - Control	Strong R	-4.096204	2.428275	783	-1.686878	0.0920251

pA_H4a <- data_clean |>
  filter(!is.na(exp_group_category)) |>
  group_by(party_numeric, exp_group_category) |>
  summarise(
    mean_y = mean(ft_rep_post, na.rm = TRUE),
    se_y   = sd(ft_rep_post, na.rm = TRUE) / sqrt(n()),
    .groups = "drop"
  ) |>
  ggplot(aes(x = factor(party_numeric), y = mean_y, color = exp_group_category)) +
  geom_errorbar(aes(ymin = mean_y - 1.96 * se_y,
                    ymax = mean_y + 1.96 * se_y),
                width = 0.15, linewidth = 0.6,
                position = position_dodge(width = 0.4)) +
  geom_point(size = 3, position = position_dodge(width = 0.4)) +
  scale_x_discrete(labels = pid_labels) +
  scale_color_manual(
    values = c(
      "Control R"   = "darkred",
      "Treatment R" = "hotpink",
      "Control D"   = "midnightblue",
      "Treatment D" = "dodgerblue"
    ),
    name = "Condition"
  ) +
  labs(title = "A", x = NULL,
       y = "Affective Ratings:\nRepublican Party\n(raw mean)") +
  base_theme +
  theme(legend.position = "bottom")

pB_H4a <- ggplot() +
  geom_hline(yintercept = 0, linetype = "dashed", color = "grey50") +
  geom_smooth(data = contr_H4a_factor_df,
              aes(x = party_numeric, y = estimate),
              method = "lm_robust", se = TRUE,
              color = "grey20", fill = "grey60",
              alpha = 0.2, linewidth = 0.9) +
  geom_errorbar(data = contr_H4a_factor_df,
                aes(x = party_numeric, ymin = lower.CL, ymax = upper.CL,
                    color = factor(party_numeric)),
                width = 0.15, linewidth = 0.6) +
  geom_point(data = contr_H4a_factor_df,
             aes(x = party_numeric, y = estimate,
                 color = factor(party_numeric)),
             size = 3) +
  scale_color_manual(values = pid_colors, guide = "none") +
  scale_x_continuous(breaks = pid_breaks, labels = pid_labels) +
  labs(title = "B", x = "Party ID",
       y = "Difference in Means") +
  base_theme

Hypothesis 4b

emm_H4b<- emmeans(H4b, ~ condition_dummy | party_numeric,
                 at = list(party_numeric = c(1, 2, 3, 5, 6, 7)),
                 vcov. = vcov(H4b))
contr_H4b<- contrast(emm_H4b,
                    method = list("Treatment - Control" = c(-1, 1)),
                    by = "party_numeric")
contr_H4b_df <- as.data.frame(confint(contr_H4b))
contr_H4b_table <- summary(contr_H4b) |>
  as.data.frame() |>
  dplyr::select(contrast, party_numeric, estimate, SE, df, t.ratio, p.value) |>
  mutate(party_numeric = factor(party_numeric, levels = c(1, 2, 3, 5, 6, 7),
                                labels = c("Strong D","Weak D","Lean D","Lean R","Weak R","Strong R")))

contrast	party_numeric	estimate	SE	df	t.ratio	p.value
Treatment - Control	Strong D	-2.7470714	1.1668581	780	-2.3542463	0.0188074
Treatment - Control	Weak D	-2.4389069	0.7780141	780	-3.1347848	0.0017842
Treatment - Control	Lean D	-2.1307424	0.5014888	780	-4.2488337	0.0000241
Treatment - Control	Lean R	-1.5144135	0.8667016	780	-1.7473297	0.0809737
Treatment - Control	Weak R	-1.2062490	1.2667547	780	-0.9522357	0.3412724
Treatment - Control	Strong R	-0.8980846	1.6913308	780	-0.5309929	0.5955749

pA_H4b <- data_clean |>
  filter(!is.na(exp_group_category)) |>
  group_by(party_numeric, exp_group_category) |>
  summarise(
    mean_y = mean(ft_rep_pre, na.rm = TRUE),
    se_y   = sd(ft_rep_pre, na.rm = TRUE) / sqrt(n()),
    .groups = "drop"
  ) |>
  ggplot(aes(x = factor(party_numeric), y = mean_y, color = exp_group_category)) +
  geom_errorbar(aes(ymin = mean_y - 1.96 * se_y,
                    ymax = mean_y + 1.96 * se_y),
                width = 0.15, linewidth = 0.6,
                position = position_dodge(width = 0.4)) +
  geom_point(size = 3, position = position_dodge(width = 0.4)) +
  scale_x_discrete(labels = pid_labels) +
  scale_color_manual(
    values = c(
      "Control R"   = "darkred",
      "Treatment R" = "hotpink",
      "Control D"   = "midnightblue",
      "Treatment D" = "dodgerblue"
    ),
    name = "Condition"
  ) +
  labs(title = "A", x = NULL,
       y = "Affective Ratings:\nRepublican Party\n(pre-treatment mean)") +
  base_theme +
  theme(legend.position = "bottom")

pB_H4b <- ggplot() +
  geom_hline(yintercept = 0, linetype = "dashed", color = "grey50") +
  geom_smooth(data = contr_H4b_factor_df,
              aes(x = party_numeric, y = estimate),
              method = "lm_robust", se = TRUE,
              color = "grey20", fill = "grey60",
              alpha = 0.2, linewidth = 0.9) +
  geom_errorbar(data = contr_H4b_factor_df,
                aes(x = party_numeric, ymin = lower.CL, ymax = upper.CL,
                    color = factor(party_numeric)),
                width = 0.15, linewidth = 0.6) +
  geom_point(data = contr_H4b_factor_df,
             aes(x = party_numeric, y = estimate,
                 color = factor(party_numeric)),
             size = 3) +
  scale_color_manual(values = pid_colors, guide = "none") +
  scale_x_continuous(breaks = pid_breaks, labels = pid_labels) +
  labs(title = "B", x = "Party ID",
       y = "Difference in Means") +
  base_theme

Hypothesis 5

For Hypotheses 5a through 6c, I establish three sets of planned contrasts based on partisan self-concept: weaker-identifying Republicans (H5a and H6a), stronger-identifying Republicans (H5b and H6b), and stronger-identifying Democrats (H5c and H6c). I created these bins based on the 7-point party identification scale detailed above. Given that the six groups referenced are “Strong D,” “Weak D”, “Lean D”, “Lean R”, “Weak R,” and “Strong R”, I plan to focus on the results for “Lean R”, “Strong D,” and “Strong R” as our primary groups of interest for each of the respective hypotheses.

emm_H5 <- emmeans(H5, ~ condition_dummy | party_numeric + PID_bin,
                       at = list(party_numeric = c(1, 2, 3, 5, 6, 7), PID_bin = c(0, 1, 2)),
                       vcov. = vcov(H5))
contr_H5 <- contrast(emm_H5,
            method = list("Treatment - Control" = c(-1, 1)),
            by = c("party_numeric", "PID_bin"))
contr_table_H5 <- summary(contr_H5) |>
  as.data.frame() |>
          dplyr::select(contrast, party_numeric, PID_bin, estimate, SE, df, t.ratio, p.value) |>
          mutate(
            party_numeric = factor(party_numeric, levels = c(1, 2, 3, 5, 6, 7),
                                   labels = c("Strong D", "Weak D", "Lean D",
                                              "Lean R", "Weak R", "Strong R")),
            PID_bin = factor(PID_bin, levels = 0:2,
                             labels = c("Weak", "Medium", "Strong")))

contrast	party_numeric	PID_bin	estimate	SE	df	t.ratio	p.value
Treatment - Control	Strong D	Weak	-9.130332	6.022727	775	-1.5159797	0.1299322
Treatment - Control	Weak D	Weak	-8.081846	4.551179	775	-1.7757699	0.0761632
Treatment - Control	Lean D	Weak	-7.033361	3.696906	775	-1.9024991	0.0574763
Treatment - Control	Lean R	Weak	-4.936390	5.011645	775	-0.9849838	0.3249394
Treatment - Control	Weak R	Weak	-3.887904	6.603952	775	-0.5887239	0.5562181
Treatment - Control	Strong R	Weak	-2.839418	8.403670	775	-0.3378784	0.7355463
Treatment - Control	Strong D	Medium	-6.776571	2.978819	775	-2.2749190	0.0231832
Treatment - Control	Weak D	Medium	-6.073276	2.262054	775	-2.6848501	0.0074113
Treatment - Control	Lean D	Medium	-5.369981	1.830555	775	-2.9335263	0.0034502
Treatment - Control	Lean R	Medium	-3.963392	2.406359	775	-1.6470494	0.0999533
Treatment - Control	Weak R	Medium	-3.260097	3.161669	775	-1.0311317	0.3028008
Treatment - Control	Strong R	Medium	-2.556802	4.023959	775	-0.6353945	0.5253587
Treatment - Control	Strong D	Strong	-4.422810	2.290807	775	-1.9306776	0.0538874
Treatment - Control	Weak D	Strong	-4.064706	1.931356	775	-2.1045872	0.0356487
Treatment - Control	Lean D	Strong	-3.706602	1.702886	775	-2.1766592	0.0298072
Treatment - Control	Lean R	Strong	-2.990393	1.816931	775	-1.6458485	0.1002004
Treatment - Control	Weak R	Strong	-2.632289	2.129078	775	-1.2363519	0.2167023
Treatment - Control	Strong R	Strong	-2.274185	2.540089	775	-0.8953170	0.3708957

Weak/Med Partisan Self-Concept

p_H5_weakmed <- make_H5_H6_panels(
  "ft_rep_post", "Weak-Medium", contr_H5_factor_df,
  "Affective Ratings:\nRepublican Party\n(raw mean)",
  "Figure 4.5a: Republican Party Ratings — Weak-Medium PID Strength"
)
p_H5_weakmed

Strong Partisan Self-Concept

p_H5_strong <- make_H5_H6_panels(
  "ft_rep_post", "Strong", contr_H5_factor_df,
  "Affective Ratings:\nRepublican Party\n(raw mean)",
  "Figure 4.5b: Republican Party Ratings — Strong PID Strength"
)
p_H5_strong

Hypothesis 6

emm_H6 <- emmeans(H6, ~ condition_dummy | party_numeric + PID_bin,
                       at = list(party_numeric = c(1, 2, 3, 5, 6, 7), PID_bin = c(0, 1, 2)),
                       vcov. = vcov(H6))
contr_H6 <- contrast(emm_H6,
            method = list("Treatment - Control" = c(-1, 1)),
            by = c("party_numeric", "PID_bin"))
contr_table_H6 <- summary(contr_H6) |>
  as.data.frame() |>
          dplyr::select(contrast, party_numeric, PID_bin, estimate, SE, df, t.ratio, p.value) |>
          mutate(
            party_numeric = factor(party_numeric, levels = c(1, 2, 3, 5, 6, 7),
                                   labels = c("Strong D", "Weak D", "Lean D",
                                              "Lean R", "Weak R", "Strong R")),
            PID_bin = factor(PID_bin, levels = 0:2,
                             labels = c("Weak", "Medium", "Strong")))

contrast	party_numeric	PID_bin	estimate	SE	df	t.ratio	p.value
Treatment - Control	Strong D	Weak	-0.0856868	7.348324	775	-0.0116607	0.9906993
Treatment - Control	Weak D	Weak	1.3358716	5.256455	775	0.2541393	0.7994554
Treatment - Control	Lean D	Weak	2.7574300	4.191687	775	0.6578329	0.5108408
Treatment - Control	Lean R	Weak	5.6005468	6.809738	775	0.8224321	0.4110840
Treatment - Control	Weak R	Weak	7.0221052	9.232274	775	0.7606041	0.4471249
Treatment - Control	Strong R	Weak	8.4436637	11.849250	775	0.7125906	0.4763136
Treatment - Control	Strong D	Medium	0.2947229	3.597686	775	0.0819201	0.9347314
Treatment - Control	Weak D	Medium	0.5656399	2.583925	775	0.2189073	0.8267799
Treatment - Control	Lean D	Medium	0.8365569	2.064616	775	0.4051877	0.6854514
Treatment - Control	Lean R	Medium	1.3783909	3.315501	775	0.4157414	0.6777144
Treatment - Control	Weak R	Medium	1.6493078	4.487369	775	0.3675445	0.7133132
Treatment - Control	Strong R	Medium	1.9202248	5.756133	775	0.3335963	0.7387744
Treatment - Control	Strong D	Strong	0.6751326	1.781732	775	0.3789193	0.7048516
Treatment - Control	Weak D	Strong	-0.2045918	1.544002	775	-0.1325075	0.8946173
Treatment - Control	Lean D	Strong	-1.0843163	1.701438	775	-0.6372940	0.5241216
Treatment - Control	Lean R	Strong	-2.8437651	2.796697	775	-1.0168299	0.3095517
Treatment - Control	Weak R	Strong	-3.7234896	3.498169	775	-1.0644109	0.2874741
Treatment - Control	Strong R	Strong	-4.6032140	4.237268	775	-1.0863637	0.2776559

Weak/Med Partisan Self-Concept

p_H6_weakmed <- make_H5_H6_panels(
  "ft_dem_post", "Weak-Medium", contr_H6_factor_df,
  "Affective Ratings:\nDemocratic Party\n(raw mean)",
  "Figure 4.6a: Democratic Party Ratings — Weak-Medium PID Strength"
)
p_H6_weakmed

Strong Partisan Self-Concept

p_H6_strong <- make_H5_H6_panels(
  "ft_dem_post", "Strong", contr_H6_factor_df,
  "Affective Ratings:\nDemocratic Party\n(raw mean)",
  "Figure 4.6b: Democratic Party Ratings — Strong PID Strength"
)
p_H6_strong

Exploratory Analyses

Post-Treatment FT as Outcome Variable

In addition to the pre-planned analyses, I will perform exploratory analyses on the post-treatment affective ratings for both parties, as I found during the pilot study that this measure was the only significant variable across all hypothesis models. I will first employ two Ordinary Least Squares (OLS) linear regression models with heteroskedasticity-robust standard errors, one for each party, regressing the party-specific post-treatment feeling thermometer ratings onto a three-way interaction between the treatment, partisan identity, and party-specific pre-treatment feeling thermometer rating as a continuous moderator. No other covariates will be included at outset.

model_ft_dem <- lm_robust(ft_dem_post ~ condition_dummy*PID_dummy*ft_dem_pre, data = data_clean)
model_ft_rep <- lm_robust(ft_rep_post ~ condition_dummy*PID_dummy*ft_rep_pre, data = data_clean)

FT Democrat

term	estimate	std.error	statistic	p.value	conf.low	conf.high	df	outcome
(Intercept)	1.129	1.335	0.846	0.398	-1.491	3.749	777	ft_dem_post
condition_dummy	0.167	1.699	0.098	0.922	-3.168	3.503	777	ft_dem_post
PID_dummy	-0.088	1.500	-0.059	0.953	-3.034	2.857	777	ft_dem_post
ft_dem_pre	0.983	0.016	60.641	0.000	0.951	1.015	777	ft_dem_post
condition_dummy:PID_dummy	1.126	2.260	0.498	0.618	-3.309	5.562	777	ft_dem_post
condition_dummy:ft_dem_pre	0.005	0.021	0.240	0.810	-0.036	0.046	777	ft_dem_post
PID_dummy:ft_dem_pre	-0.017	0.025	-0.686	0.493	-0.066	0.032	777	ft_dem_post
condition_dummy:PID_dummy:ft_dem_pre	-0.056	0.052	-1.075	0.283	-0.159	0.046	777	ft_dem_post

FT Republican

term	estimate	std.error	statistic	p.value	conf.low	conf.high	df	outcome
(Intercept)	0.972	0.562	1.730	0.084	-0.131	2.075	778	ft_rep_post
condition_dummy	-1.569	0.685	-2.290	0.022	-2.913	-0.224	778	ft_rep_post
PID_dummy	2.611	2.675	0.976	0.329	-2.641	7.862	778	ft_rep_post
ft_rep_pre	0.946	0.026	36.540	0.000	0.895	0.996	778	ft_rep_post
condition_dummy:PID_dummy	0.975	3.403	0.287	0.774	-5.704	7.654	778	ft_rep_post
condition_dummy:ft_rep_pre	-0.030	0.039	-0.769	0.442	-0.107	0.047	778	ft_rep_post
PID_dummy:ft_rep_pre	0.004	0.041	0.086	0.932	-0.077	0.084	778	ft_rep_post
condition_dummy:PID_dummy:ft_rep_pre	0.014	0.060	0.236	0.813	-0.103	0.132	778	ft_rep_post

# |--------------------------------|
#         EXP_GROUP_CATEGORY
# |--------------------------------|
# 1.1: plot for Democratic Party FT
ggplot(data_clean, aes(x = ft_dem_pre, y = ft_dem_post,
                       color = exp_group_category, fill = exp_group_category, group = exp_group_category)) +
  geom_point() + 
  geom_smooth(method = "lm_robust", se = TRUE) +
  scale_color_manual(name = "Group", values = c("firebrick", "midnightblue","violetred", "dodgerblue"),
                     labels = c("Control (R)", "Control (D)", "Treatment (R)", "Treatment (D)")) +
  scale_fill_manual(name = "Group", values = c("lightsalmon", "lightsteelblue","lightpink", "lightblue"),
                    labels = c("Control (R)", "Control (D)", "Treatment (R)", "Treatment (D)")) +
  labs(x = "Pre-Treatment Ratings",
       y = "Post-Treatment Ratings",
       title = "Post-Treatment Affective Ratings for Democratic Party By Pre-Treatment Affective Ratings") +
  theme_minimal()

# 1.2: plot for Republican Party FT
ggplot(data_clean, aes(x = ft_rep_pre, y = ft_rep_post,
                       color = exp_group_category, fill = exp_group_category, group = exp_group_category)) +
  geom_point() +
  geom_smooth(method = "lm_robust", se = TRUE) +
  scale_color_manual(name = "Group", values = c("firebrick", "midnightblue","violetred", "dodgerblue"),
                     labels = c("Control (R)", "Control (D)", "Treatment (R)", "Treatment (D)")) +
  scale_fill_manual(name = "Group", values = c("lightsalmon", "lightsteelblue","lightpink", "lightblue"),
                    labels = c("Control (R)", "Control (D)", "Treatment (R)", "Treatment (D)")) +
  labs(x = "Pre-Treatment Ratings",
       y = "Post-Treatment Ratings",
       title = "Post-Treatment Affective Ratings for Republican Party By Pre-Treatment Affective Ratings") +
  theme_minimal()

# |--------------------------------|
#            PARTY_BIN
# |--------------------------------|
# 2.1: plot for Democratic Party FT
ggplot(data_clean, aes(x = ft_dem_pre, y = ft_dem_post,
                       color = party_category, fill = party_category, group = party_category)) +
  geom_point() + 
  geom_smooth(method = "lm_robust", se = TRUE) +
  scale_color_manual(name = "Party ID", values = c("orchid", "dodgerblue", "firebrick"),
                     labels = c("Independent", "Democrat", "Republican")) +
  scale_fill_manual(name = "Party ID", values = c("thistle", "lightblue", "lightpink"),
                    labels = c("Independent", "Democrat", "Republican")) +
  labs(x = "Pre-Treatment Ratings",
       y = "Post-Treatment Ratings",
       title = "Post-Treatment Affective Ratings for Democratic Party By Party ID") +
  theme_minimal()

# 2.2: plot for Democratic Party FT   
ggplot(data_clean, aes(x = ft_rep_pre, y = ft_rep_post,
                       color = party_category, fill = party_category, group = party_category)) +
  geom_point() +
  geom_smooth(method = "lm_robust", se = TRUE) +
  scale_color_manual(name = "Party ID", values = c("orchid", "dodgerblue", "firebrick"),
                     labels = c("Independent", "Democrat", "Republican")) +
  scale_fill_manual(name = "Party ID", values = c("thistle", "lightblue", "lightpink"),
                    labels = c("Independent", "Democrat", "Republican")) +
  labs(x = "Pre-Treatment Ratings",
       y = "Post-Treatment Ratings",
       title = "Post-Treatment Affective Ratings for Republican Party By Party ID") +
  theme_minimal()

I will then craft separate exploratory models adding the following covariates individually: PID strength, MAGA support, ideology, and the expanded party scale. Given that I focus on both affective ratings for the Democratic and Republican Party post-treatment, I will produce two models per individual covariate, i.e., two models for PID strength, two for MAGA support, etc. For the expanded party scale, I will replace the dichotomous variable for partisan identification with the 7-point ANES scale, producing a three-way interaction between the treatment, the expanded party scale, and the party-specific pre-treatment feeling thermometer rating. For PID strength, MAGA support, and ideology, I will add a three-way interaction between the treatment, the dichotomous measure of partisanship, and the chosen covariate to the initial model described.

Partisan Self-Concept

Democrats

lm_dem_pidstr <- lm_robust(ft_dem_post ~ condition_dummy*PID_dummy*ft_dem_pre
                           + condition_dummy*PID_dummy*PID_strength,
                           data = data_clean, se_type = "HC2")

FT Democrat

term	estimate	std.error	statistic	p.value	conf.low	conf.high	df	outcome
(Intercept)	0.311	1.148	0.271	0.787	-1.943	2.564	769	ft_dem_post
condition_dummy	0.317	1.587	0.200	0.842	-2.799	3.433	769	ft_dem_post
PID_dummy	-0.065	1.793	-0.036	0.971	-3.584	3.454	769	ft_dem_post
ft_dem_pre	0.950	0.033	28.680	0.000	0.885	1.015	769	ft_dem_post
PID_strength	1.026	0.699	1.468	0.143	-0.346	2.398	769	ft_dem_post
condition_dummy:PID_dummy	6.639	4.735	1.402	0.161	-2.656	15.935	769	ft_dem_post
condition_dummy:ft_dem_pre	0.007	0.041	0.176	0.860	-0.073	0.088	769	ft_dem_post
PID_dummy:ft_dem_pre	0.021	0.038	0.547	0.584	-0.053	0.095	769	ft_dem_post
condition_dummy:PID_strength	-0.112	0.942	-0.119	0.905	-1.962	1.738	769	ft_dem_post
PID_dummy:PID_strength	-0.889	0.849	-1.047	0.296	-2.557	0.778	769	ft_dem_post
condition_dummy:PID_dummy:ft_dem_pre	-0.073	0.065	-1.122	0.262	-0.201	0.055	769	ft_dem_post
condition_dummy:PID_dummy:PID_strength	-1.487	1.567	-0.949	0.343	-4.563	1.590	769	ft_dem_post

make_pA(data_clean, "PID_strength", "Partisan Self-Concept\n(raw mean)") /
  make_pB(contr_ft_D_df, "FT: Democratic Party\nDiff in Means (Treatment - Control)") +
  plot_annotation(
    title = "Figure: Post-Treatment FT (Democratic Party) by Party ID | Covariate: Partisan Self-Concept")

Republicans

lm_rep_pidstr <- lm_robust(ft_rep_post ~ condition_dummy*PID_dummy*ft_rep_pre
                           + condition_dummy*PID_dummy*PID_strength,
                           data = data_clean, se_type = "HC2")

FT Republican

term	estimate	std.error	statistic	p.value	conf.low	conf.high	df	outcome
(Intercept)	1.868	1.451	1.287	0.198	-0.981	4.716	770	ft_rep_post
condition_dummy	-1.504	1.983	-0.759	0.448	-5.396	2.388	770	ft_rep_post
PID_dummy	0.207	2.673	0.077	0.938	-5.040	5.453	770	ft_rep_post
ft_rep_pre	0.944	0.027	35.127	0.000	0.891	0.997	770	ft_rep_post
PID_strength	-0.281	0.345	-0.815	0.415	-0.959	0.397	770	ft_rep_post
condition_dummy:PID_dummy	1.230	3.854	0.319	0.750	-6.335	8.796	770	ft_rep_post
condition_dummy:ft_rep_pre	-0.030	0.040	-0.735	0.462	-0.108	0.049	770	ft_rep_post
PID_dummy:ft_rep_pre	-0.036	0.054	-0.658	0.511	-0.143	0.071	770	ft_rep_post
condition_dummy:PID_strength	-0.022	0.510	-0.042	0.966	-1.024	0.980	770	ft_rep_post
PID_dummy:PID_strength	1.788	0.824	2.168	0.030	0.169	3.406	770	ft_rep_post
condition_dummy:PID_dummy:ft_rep_pre	-0.065	0.093	-0.700	0.484	-0.246	0.117	770	ft_rep_post
condition_dummy:PID_dummy:PID_strength	1.731	1.503	1.152	0.250	-1.219	4.681	770	ft_rep_post

make_pA(data_clean, "PID_strength", "Partisan Self-Concept\n(raw mean)") /
  make_pB(contr_ft_R_df, "FT: Republican Party\nDiff in Means (Treatment - Control)") +
  plot_annotation(
    title = "Figure: Post-Treatment FT (Republican Party) by Party ID | Covariate: Partisan Self-Concept")

MAGA

Democrats

lm_maga_D <- lm_robust(ft_dem_post ~ condition_dummy*PID_dummy*ft_dem_pre
                       + maga*condition_dummy*PID_dummy, data = data_clean)

FT Democrat

term	estimate	std.error	statistic	p.value	conf.low	conf.high	df	outcome
(Intercept)	1.129	1.335	0.846	0.398	-1.491	3.749	775	ft_dem_post
condition_dummy	0.167	1.699	0.098	0.922	-3.169	3.503	775	ft_dem_post
PID_dummy	-0.554	1.618	-0.342	0.732	-3.730	2.623	775	ft_dem_post
ft_dem_pre	0.983	0.016	60.641	0.000	0.951	1.015	775	ft_dem_post
maga	NA	NA	NA	NA	NA	NA	NA	ft_dem_post
condition_dummy:PID_dummy	2.555	2.463	1.038	0.300	-2.279	7.390	775	ft_dem_post
condition_dummy:ft_dem_pre	0.005	0.021	0.240	0.810	-0.036	0.046	775	ft_dem_post
PID_dummy:ft_dem_pre	-0.013	0.025	-0.496	0.620	-0.062	0.037	775	ft_dem_post
condition_dummy:maga	-2.496	2.042	-1.223	0.222	-6.503	1.512	775	ft_dem_post
PID_dummy:maga	0.783	1.066	0.734	0.463	-1.310	2.875	775	ft_dem_post
condition_dummy:PID_dummy:ft_dem_pre	-0.073	0.053	-1.365	0.173	-0.178	0.032	775	ft_dem_post
condition_dummy:PID_dummy:maga	NA	NA	NA	NA	NA	NA	NA	ft_dem_post

make_pA(data_clean, "maga", "MAGA Support\n(raw mean)") /
  make_pB(contr_ft_D_df, "FT: Democratic Party\nDiff in Means (Treatment - Control)") +
  plot_annotation(
    title = "Figure: Post-Treatment FT (Democratic Party) by Party ID | Covariate: MAGA Support")

Republicans

lm_maga_R <- lm_robust(ft_rep_post ~ condition_dummy*PID_dummy*ft_rep_pre
                       + maga*condition_dummy*PID_dummy, data = data_clean)

FT Republican

term	estimate	std.error	statistic	p.value	conf.low	conf.high	df	outcome
(Intercept)	0.972	0.562	1.730	0.084	-0.131	2.075	776	ft_rep_post
condition_dummy	-1.569	0.685	-2.290	0.022	-2.913	-0.224	776	ft_rep_post
PID_dummy	2.491	2.633	0.946	0.345	-2.678	7.660	776	ft_rep_post
ft_rep_pre	0.946	0.026	36.540	0.000	0.895	0.996	776	ft_rep_post
maga	NA	NA	NA	NA	NA	NA	NA	ft_rep_post
condition_dummy:PID_dummy	0.433	3.267	0.133	0.895	-5.980	6.846	776	ft_rep_post
condition_dummy:ft_rep_pre	-0.030	0.039	-0.769	0.442	-0.107	0.047	776	ft_rep_post
PID_dummy:ft_rep_pre	0.009	0.041	0.219	0.827	-0.071	0.089	776	ft_rep_post
condition_dummy:maga	5.111	1.724	2.965	0.003	1.727	8.495	776	ft_rep_post
PID_dummy:maga	-0.724	0.888	-0.815	0.415	-2.466	1.019	776	ft_rep_post
condition_dummy:PID_dummy:ft_rep_pre	-0.001	0.060	-0.015	0.988	-0.118	0.117	776	ft_rep_post
condition_dummy:PID_dummy:maga	NA	NA	NA	NA	NA	NA	NA	ft_rep_post

make_pA(data_clean, "maga", "MAGA Support\n(raw mean)") /
  make_pB(contr_ft_R_df, "FT: Republican Party\nDiff in Means (Treatment - Control)") +
  plot_annotation(
    title = "Figure: Post-Treatment FT (Republican Party) by Party ID | Covariate: MAGA Support")

Ideology

Democrats

lm_ideology_selfreport_D <- lm_robust(
  ft_dem_post ~ condition_dummy*PID_dummy*ft_dem_pre
  + condition_dummy*PID_dummy*ideology_selfreport,
  data = data_clean)

FT Democrat

term	estimate	std.error	statistic	p.value	conf.low	conf.high	df	outcome
(Intercept)	0.396	1.377	0.287	0.774	-2.307	3.099	773	ft_dem_post
condition_dummy	2.330	2.021	1.153	0.249	-1.637	6.297	773	ft_dem_post
PID_dummy	0.945	3.390	0.279	0.781	-5.710	7.600	773	ft_dem_post
ft_dem_pre	0.985	0.015	65.806	0.000	0.955	1.014	773	ft_dem_post
ideology_selfreport	0.273	0.551	0.495	0.621	-0.810	1.355	773	ft_dem_post
condition_dummy:PID_dummy	13.045	7.857	1.660	0.097	-2.379	28.469	773	ft_dem_post
condition_dummy:ft_dem_pre	0.003	0.020	0.143	0.886	-0.036	0.042	773	ft_dem_post
PID_dummy:ft_dem_pre	-0.019	0.024	-0.817	0.414	-0.066	0.027	773	ft_dem_post
condition_dummy:ideology_selfreport	-0.973	0.747	-1.302	0.193	-2.439	0.494	773	ft_dem_post
PID_dummy:ideology_selfreport	-0.323	0.755	-0.427	0.669	-1.805	1.160	773	ft_dem_post
condition_dummy:PID_dummy:ft_dem_pre	-0.087	0.058	-1.500	0.134	-0.200	0.027	773	ft_dem_post
condition_dummy:PID_dummy:ideology_selfreport	-1.343	1.377	-0.976	0.330	-4.047	1.360	773	ft_dem_post

make_pA(data_clean, "ideology_selfreport", "Self-Reported Ideology\n(raw mean)") /
  make_pB(contr_ft_D_df, "FT: Democratic Party\nDiff in Means (Treatment - Control)") +
  plot_annotation(
    title = "Figure: Post-Treatment FT (Democratic Party) by Party ID | Covariate: Ideology")

Republicans

lm_ideology_selfreport_R <- lm_robust(
  ft_rep_post ~ condition_dummy*PID_dummy*ft_rep_pre
  + condition_dummy*PID_dummy*ideology_selfreport,
  data = data_clean)

FT Republican

term	estimate	std.error	statistic	p.value	conf.low	conf.high	df	outcome
(Intercept)	-0.392	0.544	-0.722	0.470	-1.459	0.675	774	ft_rep_post
condition_dummy	1.094	1.034	1.057	0.291	-0.937	3.125	774	ft_rep_post
PID_dummy	0.889	3.518	0.253	0.801	-6.018	7.795	774	ft_rep_post
ft_rep_pre	0.933	0.030	30.961	0.000	0.874	0.992	774	ft_rep_post
ideology_selfreport	0.729	0.436	1.672	0.095	-0.127	1.585	774	ft_rep_post
condition_dummy:PID_dummy	-4.989	5.162	-0.967	0.334	-15.123	5.144	774	ft_rep_post
condition_dummy:ft_rep_pre	-0.002	0.043	-0.035	0.972	-0.087	0.084	774	ft_rep_post
PID_dummy:ft_rep_pre	0.003	0.047	0.074	0.941	-0.088	0.095	774	ft_rep_post
condition_dummy:ideology_selfreport	-1.528	0.673	-2.269	0.024	-2.850	-0.206	774	ft_rep_post
PID_dummy:ideology_selfreport	0.003	0.748	0.003	0.997	-1.467	1.472	774	ft_rep_post
condition_dummy:PID_dummy:ft_rep_pre	-0.027	0.068	-0.393	0.694	-0.161	0.107	774	ft_rep_post
condition_dummy:PID_dummy:ideology_selfreport	2.276	1.271	1.791	0.074	-0.219	4.772	774	ft_rep_post

make_pA(data_clean, "ideology_selfreport", "Self-Reported Ideology\n(raw mean)") /
  make_pB(contr_ft_R_df, "FT: Republican Party\nDiff in Means (Treatment - Control)") +
  plot_annotation(
    title = "Figure: Post-Treatment FT (Republican Party) by Party ID | Covariate: Ideology")

Party ID Strength (7-pt)

Democrats

lm_party_numeric_D <- lm_robust(
  ft_dem_post ~ condition_dummy*party_numeric*ft_dem_pre,
  data = data_clean)

FT Democrat

term	estimate	std.error	statistic	p.value	conf.low	conf.high	df	outcome
(Intercept)	1.024	2.044	0.501	0.616	-2.987	5.036	777	ft_dem_post
condition_dummy	2.763	2.761	1.001	0.317	-2.656	8.182	777	ft_dem_post
party_numeric	0.041	0.365	0.112	0.911	-0.676	0.758	777	ft_dem_post
ft_dem_pre	0.993	0.024	42.039	0.000	0.946	1.039	777	ft_dem_post
condition_dummy:party_numeric	-0.317	0.523	-0.607	0.544	-1.343	0.708	777	ft_dem_post
condition_dummy:ft_dem_pre	-0.014	0.032	-0.446	0.655	-0.078	0.049	777	ft_dem_post
party_numeric:ft_dem_pre	-0.006	0.005	-1.137	0.256	-0.015	0.004	777	ft_dem_post
condition_dummy:party_numeric:ft_dem_pre	-0.005	0.010	-0.498	0.618	-0.025	0.015	777	ft_dem_post

make_pA(data_clean, "ft_dem_pre", "Pre-Treatment FT: Democratic Party\n(raw mean)") /
  make_pB(contr_ft_D_df, "FT: Democratic Party\nDiff in Means (Treatment - Control)") +
  plot_annotation(
    title = "Figure: Post-Treatment FT (Democratic Party) by Party Affiliation Strength")

Republicans

lm_party_numeric_R <- lm_robust(
  ft_rep_post ~ condition_dummy*party_numeric*ft_rep_pre,
  data = data_clean)

FT Republican

term	estimate	std.error	statistic	p.value	conf.low	conf.high	df	outcome
(Intercept)	-0.326	0.792	-0.411	0.681	-1.880	1.228	778	ft_rep_post
condition_dummy	-0.840	1.026	-0.819	0.413	-2.853	1.173	778	ft_rep_post
party_numeric	0.865	0.474	1.823	0.069	-0.066	1.796	778	ft_rep_post
ft_rep_pre	0.936	0.034	27.652	0.000	0.869	1.002	778	ft_rep_post
condition_dummy:party_numeric	-0.307	0.565	-0.543	0.587	-1.417	0.802	778	ft_rep_post
condition_dummy:ft_rep_pre	-0.061	0.052	-1.170	0.242	-0.163	0.041	778	ft_rep_post
party_numeric:ft_rep_pre	-0.001	0.007	-0.087	0.931	-0.014	0.013	778	ft_rep_post
condition_dummy:party_numeric:ft_rep_pre	0.011	0.010	1.181	0.238	-0.008	0.031	778	ft_rep_post

make_pA(data_clean, "ft_rep_pre", "Pre-Treatment FT: Republican Party\n(raw mean)") /
  make_pB(contr_ft_R_df, "FT: Republican Party\nDiff in Means (Treatment - Control)") +
  plot_annotation(
    title = "Figure: Post-Treatment FT (Republican Party) by Party Affiliation Strength")

Social Closeness

I deviate from the pre-analysis plan by also including social closeness as a covariate against the same outcome variables and main predictor variables.

Democrats

lm_social_closeness_D <- lm_robust(
  ft_dem_post ~ condition_dummy*PID_dummy*ft_dem_pre
  + condition_dummy*PID_dummy*social_closeness,
  data = data_clean)

FT Democrat

term	estimate	std.error	statistic	p.value	conf.low	conf.high	df	outcome
(Intercept)	1.324	1.601	0.827	0.409	-1.820	4.468	771	ft_dem_post
condition_dummy	2.778	2.321	1.197	0.232	-1.779	7.335	771	ft_dem_post
PID_dummy	-4.971	2.641	-1.882	0.060	-10.155	0.213	771	ft_dem_post
ft_dem_pre	0.983	0.016	61.885	0.000	0.952	1.014	771	ft_dem_post
social_closeness	-0.074	0.613	-0.122	0.903	-1.277	1.128	771	ft_dem_post
condition_dummy:PID_dummy	-2.933	4.917	-0.597	0.551	-12.585	6.718	771	ft_dem_post
condition_dummy:ft_dem_pre	0.011	0.021	0.544	0.587	-0.029	0.052	771	ft_dem_post
PID_dummy:ft_dem_pre	-0.047	0.029	-1.600	0.110	-0.105	0.011	771	ft_dem_post
condition_dummy:social_closeness	-1.259	0.780	-1.615	0.107	-2.790	0.272	771	ft_dem_post
PID_dummy:social_closeness	2.013	1.038	1.939	0.053	-0.025	4.051	771	ft_dem_post
condition_dummy:PID_dummy:ft_dem_pre	-0.059	0.066	-0.889	0.374	-0.188	0.071	771	ft_dem_post
condition_dummy:PID_dummy:social_closeness	1.547	1.938	0.798	0.425	-2.258	5.351	771	ft_dem_post

make_pA(data_clean, "social_closeness", "Social Closeness\n(raw mean)") /
  make_pB(contr_ft_D_df, "FT: Democratic Party\nDiff in Means (Treatment - Control)") +
  plot_annotation(
    title = "Figure: Post-Treatment FT (Democratic Party) by Party ID | Covariate: Social Closeness")

Republicans

lm_social_closeness_R <- lm_robust(
  ft_rep_post ~ condition_dummy*PID_dummy*ft_rep_pre
  + condition_dummy*PID_dummy*social_closeness,
  data = data_clean)

FT Republican

term	estimate	std.error	statistic	p.value	conf.low	conf.high	df	outcome
(Intercept)	-2.718	1.535	-1.770	0.077	-5.732	0.296	772	ft_rep_post
condition_dummy	0.192	1.728	0.111	0.912	-3.201	3.585	772	ft_rep_post
PID_dummy	7.119	4.366	1.630	0.103	-1.452	15.690	772	ft_rep_post
ft_rep_pre	0.899	0.038	23.439	0.000	0.823	0.974	772	ft_rep_post
social_closeness	1.894	0.810	2.338	0.020	0.304	3.483	772	ft_rep_post
condition_dummy:PID_dummy	9.201	6.995	1.315	0.189	-4.531	22.932	772	ft_rep_post
condition_dummy:ft_rep_pre	-0.007	0.053	-0.126	0.900	-0.111	0.098	772	ft_rep_post
PID_dummy:ft_rep_pre	0.050	0.050	0.997	0.319	-0.049	0.149	772	ft_rep_post
condition_dummy:social_closeness	-0.900	0.924	-0.974	0.330	-2.714	0.914	772	ft_rep_post
PID_dummy:social_closeness	-2.157	1.072	-2.011	0.045	-4.262	-0.052	772	ft_rep_post
condition_dummy:PID_dummy:ft_rep_pre	-0.027	0.072	-0.371	0.711	-0.167	0.114	772	ft_rep_post
condition_dummy:PID_dummy:social_closeness	-1.822	1.692	-1.077	0.282	-5.142	1.499	772	ft_rep_post

make_pA(data_clean, "social_closeness", "Social Closeness\n(raw mean)") /
  make_pB(contr_ft_R_df, "FT: Republican Party\nDiff in Means (Treatment - Control)") +
  plot_annotation(
    title = "Figure: Post-Treatment FT (Republican Party) by Party ID | Covariate: Social Closeness")

Hypotheses Tests With Covariates

Additionally, I outline several exploratory analyses focused on support for political violence within each defined scenario. Primarily, I will doing an exploratory test of all included measures related to participants’ political “identification”, i.e., partisan identification, ideology, partisan self-concept and social closeness. I will craft a second Ordinary Least Squares (OLS) linear regression models with heteroskedasticity-robust standard errors for each of the 14 proposed hypotheses previously outlined to determine whether the introduction of any additional pre-treatment covariates alters the significance of the treatment effect. I have selected support for the MAGA movement, self-reported ideology, partisan self-concept, and social closeness as our socio-political variables of interest across all models, with ideology and MAGA support as the only two variables which I do not interact with the treatment given that the others correspond more strongly to partisanship as a social and political identity rather than a labeled affiliation.

Hypothesis 1

H1_cov <- lm_robust(SPV_meta_R ~ condition_dummy * factor(party_numeric)
                    + condition_dummy * PID_strength
                    + condition_dummy * social_closeness
                    + ideology_selfreport
                    + maga,
                    data = data_clean, se_type = "HC2")

term	estimate	std.error	statistic	p.value	conf.low	conf.high	df	outcome
(Intercept)	48.881	7.553	6.472	0.000	34.054	63.708	761	SPV_meta_R
condition_dummy	14.456	11.918	1.213	0.226	-8.941	37.853	761	SPV_meta_R
factor(party_numeric)2	5.285	4.079	1.296	0.195	-2.722	13.292	761	SPV_meta_R
factor(party_numeric)3	9.416	5.223	1.803	0.072	-0.836	19.669	761	SPV_meta_R
factor(party_numeric)5	6.862	5.630	1.219	0.223	-4.189	17.914	761	SPV_meta_R
factor(party_numeric)6	1.675	4.882	0.343	0.732	-7.909	11.258	761	SPV_meta_R
factor(party_numeric)7	4.934	5.432	0.908	0.364	-5.730	15.599	761	SPV_meta_R
PID_strength	2.986	1.577	1.894	0.059	-0.109	6.081	761	SPV_meta_R
social_closeness	-5.817	1.592	-3.653	0.000	-8.942	-2.691	761	SPV_meta_R
ideology_selfreport	-2.352	0.936	-2.514	0.012	-4.189	-0.516	761	SPV_meta_R
maga	-5.492	2.961	-1.855	0.064	-11.305	0.321	761	SPV_meta_R
condition_dummy:factor(party_numeric)2	3.081	5.693	0.541	0.589	-8.095	14.257	761	SPV_meta_R
condition_dummy:factor(party_numeric)3	-1.896	7.368	-0.257	0.797	-16.361	12.569	761	SPV_meta_R
condition_dummy:factor(party_numeric)5	1.061	9.247	0.115	0.909	-17.092	19.215	761	SPV_meta_R
condition_dummy:factor(party_numeric)6	-2.016	6.024	-0.335	0.738	-13.841	9.810	761	SPV_meta_R
condition_dummy:factor(party_numeric)7	-0.484	5.996	-0.081	0.936	-12.254	11.287	761	SPV_meta_R
condition_dummy:PID_strength	-3.480	2.584	-1.347	0.178	-8.552	1.593	761	SPV_meta_R
condition_dummy:social_closeness	0.269	2.404	0.112	0.911	-4.450	4.988	761	SPV_meta_R

Hypothesis 2

H2_cov <- lm_robust(moral_disengagement_outgroup ~ condition_dummy * factor(party_numeric)
                    + condition_dummy * PID_strength
                    + condition_dummy * social_closeness
                    + ideology_selfreport
                    + maga,
                    data = data_clean, se_type = "HC2")

term	estimate	std.error	statistic	p.value	conf.low	conf.high	df	outcome
(Intercept)	4.727	0.266	17.759	0.000	4.205	5.250	765	moral_disengagement_outgroup
condition_dummy	0.201	0.375	0.536	0.592	-0.535	0.937	765	moral_disengagement_outgroup
factor(party_numeric)2	-0.182	0.136	-1.336	0.182	-0.450	0.086	765	moral_disengagement_outgroup
factor(party_numeric)3	-0.177	0.184	-0.963	0.336	-0.538	0.184	765	moral_disengagement_outgroup
factor(party_numeric)5	-0.634	0.235	-2.699	0.007	-1.095	-0.173	765	moral_disengagement_outgroup
factor(party_numeric)6	-0.816	0.195	-4.179	0.000	-1.199	-0.433	765	moral_disengagement_outgroup
factor(party_numeric)7	-0.802	0.213	-3.764	0.000	-1.220	-0.384	765	moral_disengagement_outgroup
PID_strength	0.077	0.059	1.310	0.191	-0.039	0.193	765	moral_disengagement_outgroup
social_closeness	-0.709	0.057	-12.413	0.000	-0.821	-0.597	765	moral_disengagement_outgroup
ideology_selfreport	0.099	0.037	2.701	0.007	0.027	0.171	765	moral_disengagement_outgroup
maga	0.390	0.127	3.071	0.002	0.140	0.639	765	moral_disengagement_outgroup
condition_dummy:factor(party_numeric)2	0.292	0.191	1.533	0.126	-0.082	0.667	765	moral_disengagement_outgroup
condition_dummy:factor(party_numeric)3	0.106	0.244	0.436	0.663	-0.372	0.585	765	moral_disengagement_outgroup
condition_dummy:factor(party_numeric)5	0.204	0.320	0.638	0.524	-0.424	0.833	765	moral_disengagement_outgroup
condition_dummy:factor(party_numeric)6	0.002	0.223	0.009	0.993	-0.436	0.440	765	moral_disengagement_outgroup
condition_dummy:factor(party_numeric)7	0.484	0.230	2.101	0.036	0.032	0.937	765	moral_disengagement_outgroup
condition_dummy:PID_strength	-0.020	0.084	-0.233	0.816	-0.184	0.145	765	moral_disengagement_outgroup
condition_dummy:social_closeness	-0.054	0.076	-0.711	0.477	-0.204	0.095	765	moral_disengagement_outgroup

Hypothesis 3

H3_cov <- lm_robust(SPV_individual ~ condition_dummy * factor(party_numeric)
                    + condition_dummy * PID_strength
                    + condition_dummy * social_closeness
                    + ideology_selfreport
                    + maga,
                    data = data_clean, se_type = "HC2")

term	estimate	std.error	statistic	p.value	conf.low	conf.high	df	outcome
(Intercept)	28.258	6.655	4.246	0.000	15.193	41.324	755	SPV_individual
condition_dummy	15.895	10.536	1.509	0.132	-4.789	36.579	755	SPV_individual
factor(party_numeric)2	6.397	3.002	2.131	0.033	0.505	12.289	755	SPV_individual
factor(party_numeric)3	9.437	4.598	2.052	0.040	0.410	18.463	755	SPV_individual
factor(party_numeric)5	15.585	5.540	2.813	0.005	4.710	26.461	755	SPV_individual
factor(party_numeric)6	12.704	4.023	3.158	0.002	4.806	20.601	755	SPV_individual
factor(party_numeric)7	15.228	4.798	3.174	0.002	5.810	24.647	755	SPV_individual
PID_strength	3.508	1.550	2.263	0.024	0.465	6.552	755	SPV_individual
social_closeness	-6.049	1.252	-4.830	0.000	-8.508	-3.590	755	SPV_individual
ideology_selfreport	-2.697	0.697	-3.871	0.000	-4.065	-1.329	755	SPV_individual
maga	-0.529	2.973	-0.178	0.859	-6.366	5.308	755	SPV_individual
condition_dummy:factor(party_numeric)2	2.815	4.334	0.650	0.516	-5.693	11.324	755	SPV_individual
condition_dummy:factor(party_numeric)3	-2.377	6.078	-0.391	0.696	-14.309	9.556	755	SPV_individual
condition_dummy:factor(party_numeric)5	-3.024	8.761	-0.345	0.730	-20.223	14.174	755	SPV_individual
condition_dummy:factor(party_numeric)6	-3.586	4.940	-0.726	0.468	-13.284	6.111	755	SPV_individual
condition_dummy:factor(party_numeric)7	2.495	5.555	0.449	0.653	-8.410	13.399	755	SPV_individual
condition_dummy:PID_strength	-2.637	2.433	-1.084	0.279	-7.414	2.140	755	SPV_individual
condition_dummy:social_closeness	-2.113	1.764	-1.198	0.231	-5.575	1.349	755	SPV_individual

Beginning with Hypothesis 4a, I include the pre-treatment affective ratings for both the Republican and Democratic Party, with the former transitioning into a treatment interaction term in Hypothesis 4b and the latter transitioning into a treatment interaction term in Hypothesis 5a.

Hypothesis 4a

H4a_cov <- lm_robust(ft_rep_post ~ condition_dummy * factor(party_numeric)
                     + condition_dummy * PID_strength
                     + condition_dummy * social_closeness
                     + ideology_selfreport
                     + maga
                     + ft_rep_pre
                     + ft_dem_pre,
                     data = data_clean, se_type = "HC2")

term	estimate	std.error	statistic	p.value	conf.low	conf.high	df	outcome
(Intercept)	-3.383	1.796	-1.883	0.060	-6.909	0.143	759	ft_rep_post
condition_dummy	3.425	2.963	1.156	0.248	-2.393	9.242	759	ft_rep_post
factor(party_numeric)2	0.015	0.852	0.018	0.986	-1.657	1.688	759	ft_rep_post
factor(party_numeric)3	2.685	1.664	1.614	0.107	-0.581	5.951	759	ft_rep_post
factor(party_numeric)5	1.451	2.022	0.718	0.473	-2.519	5.421	759	ft_rep_post
factor(party_numeric)6	4.465	1.815	2.461	0.014	0.903	8.027	759	ft_rep_post
factor(party_numeric)7	5.061	2.186	2.315	0.021	0.769	9.352	759	ft_rep_post
PID_strength	0.713	0.405	1.758	0.079	-0.083	1.509	759	ft_rep_post
social_closeness	1.439	0.461	3.120	0.002	0.533	2.344	759	ft_rep_post
ideology_selfreport	0.039	0.339	0.114	0.909	-0.628	0.705	759	ft_rep_post
maga	0.895	0.833	1.074	0.283	-0.741	2.531	759	ft_rep_post
ft_rep_pre	0.903	0.020	46.134	0.000	0.865	0.942	759	ft_rep_post
ft_dem_pre	-0.016	0.013	-1.191	0.234	-0.042	0.010	759	ft_rep_post
condition_dummy:factor(party_numeric)2	0.357	1.319	0.271	0.787	-2.233	2.946	759	ft_rep_post
condition_dummy:factor(party_numeric)3	-3.028	2.167	-1.398	0.163	-7.281	1.225	759	ft_rep_post
condition_dummy:factor(party_numeric)5	-1.797	3.968	-0.453	0.651	-9.587	5.993	759	ft_rep_post
condition_dummy:factor(party_numeric)6	-0.846	1.621	-0.522	0.602	-4.029	2.337	759	ft_rep_post
condition_dummy:factor(party_numeric)7	2.379	1.439	1.654	0.099	-0.445	5.203	759	ft_rep_post
condition_dummy:PID_strength	-0.733	0.682	-1.074	0.283	-2.071	0.606	759	ft_rep_post
condition_dummy:social_closeness	-1.155	0.572	-2.019	0.044	-2.278	-0.032	759	ft_rep_post

Hypothesis 4b

H4b_cov <- lm_robust(ft_rep_post ~ condition_dummy * factor(party_numeric)
                     + condition_dummy * PID_strength
                     + condition_dummy * social_closeness
                     + condition_dummy * ft_rep_pre
                     + ideology_selfreport
                     + maga
                     + ft_dem_pre,
                     data = data_clean, se_type = "HC2")

term	estimate	std.error	statistic	p.value	conf.low	conf.high	df	outcome
(Intercept)	-3.103	1.836	-1.690	0.091	-6.707	0.502	758	ft_rep_post
condition_dummy	2.754	2.807	0.981	0.327	-2.756	8.264	758	ft_rep_post
factor(party_numeric)2	-0.088	0.881	-0.100	0.921	-1.817	1.641	758	ft_rep_post
factor(party_numeric)3	2.603	1.688	1.542	0.124	-0.711	5.916	758	ft_rep_post
factor(party_numeric)5	0.983	2.131	0.461	0.645	-3.200	5.166	758	ft_rep_post
factor(party_numeric)6	3.932	2.035	1.932	0.054	-0.064	7.927	758	ft_rep_post
factor(party_numeric)7	4.429	2.455	1.804	0.072	-0.390	9.247	758	ft_rep_post
PID_strength	0.653	0.408	1.600	0.110	-0.148	1.455	758	ft_rep_post
social_closeness	1.300	0.491	2.649	0.008	0.337	2.264	758	ft_rep_post
ft_rep_pre	0.914	0.026	34.999	0.000	0.863	0.966	758	ft_rep_post
ideology_selfreport	0.048	0.337	0.142	0.887	-0.614	0.710	758	ft_rep_post
maga	0.814	0.816	0.998	0.318	-0.787	2.416	758	ft_rep_post
ft_dem_pre	-0.015	0.013	-1.145	0.253	-0.041	0.011	758	ft_rep_post
condition_dummy:factor(party_numeric)2	0.611	1.348	0.453	0.650	-2.034	3.256	758	ft_rep_post
condition_dummy:factor(party_numeric)3	-2.818	2.182	-1.292	0.197	-7.102	1.465	758	ft_rep_post
condition_dummy:factor(party_numeric)5	-0.909	3.958	-0.230	0.818	-8.679	6.861	758	ft_rep_post
condition_dummy:factor(party_numeric)6	0.287	2.567	0.112	0.911	-4.752	5.327	758	ft_rep_post
condition_dummy:factor(party_numeric)7	3.883	3.253	1.194	0.233	-2.503	10.269	758	ft_rep_post
condition_dummy:PID_strength	-0.630	0.652	-0.966	0.334	-1.911	0.650	758	ft_rep_post
condition_dummy:social_closeness	-0.862	0.603	-1.429	0.153	-2.045	0.322	758	ft_rep_post
condition_dummy:ft_rep_pre	-0.024	0.039	-0.625	0.532	-0.100	0.052	758	ft_rep_post

Hypothesis 5

H5_cov <- lm_robust(ft_rep_post ~ condition_dummy * factor(party_numeric) * PID_bin
                    + condition_dummy * social_closeness
                    + condition_dummy * ft_rep_pre
                    + condition_dummy * ft_dem_pre
                    + ideology_selfreport
                    + maga,
                    data = data_clean, se_type = "HC2")

term	estimate	std.error	statistic	p.value	conf.low	conf.high	df	outcome
(Intercept)	-1.838	1.344	-1.367	0.172	-4.476	0.801	747	ft_rep_post
condition_dummy	1.727	2.557	0.675	0.500	-3.293	6.748	747	ft_rep_post
factor(party_numeric)2	-1.567	2.015	-0.777	0.437	-5.524	2.390	747	ft_rep_post
factor(party_numeric)3	-1.434	1.794	-0.799	0.424	-4.957	2.088	747	ft_rep_post
factor(party_numeric)5	0.152	2.400	0.063	0.949	-4.559	4.863	747	ft_rep_post
factor(party_numeric)6	1.400	2.068	0.677	0.498	-2.659	5.460	747	ft_rep_post
factor(party_numeric)7	3.231	3.042	1.062	0.288	-2.740	9.203	747	ft_rep_post
PID_bin	0.628	0.561	1.119	0.263	-0.473	1.728	747	ft_rep_post
social_closeness	1.508	0.465	3.245	0.001	0.596	2.421	747	ft_rep_post
ft_rep_pre	0.914	0.028	33.168	0.000	0.860	0.968	747	ft_rep_post
ft_dem_pre	-0.021	0.017	-1.295	0.196	-0.054	0.011	747	ft_rep_post
ideology_selfreport	0.052	0.333	0.156	0.876	-0.601	0.705	747	ft_rep_post
maga	0.694	0.809	0.858	0.391	-0.895	2.283	747	ft_rep_post
condition_dummy:factor(party_numeric)2	-0.214	3.221	-0.066	0.947	-6.537	6.110	747	ft_rep_post
condition_dummy:factor(party_numeric)3	2.070	2.983	0.694	0.488	-3.786	7.925	747	ft_rep_post
condition_dummy:factor(party_numeric)5	3.740	5.881	0.636	0.525	-7.804	15.285	747	ft_rep_post
condition_dummy:factor(party_numeric)6	-1.926	4.010	-0.480	0.631	-9.799	5.947	747	ft_rep_post
condition_dummy:factor(party_numeric)7	3.134	4.923	0.637	0.525	-6.531	12.799	747	ft_rep_post
condition_dummy:PID_bin	-0.740	1.225	-0.604	0.546	-3.145	1.664	747	ft_rep_post
factor(party_numeric)2:PID_bin	0.705	1.363	0.517	0.605	-1.971	3.380	747	ft_rep_post
factor(party_numeric)3:PID_bin	4.340	3.001	1.446	0.149	-1.552	10.232	747	ft_rep_post
factor(party_numeric)5:PID_bin	0.063	1.736	0.036	0.971	-3.345	3.471	747	ft_rep_post
factor(party_numeric)6:PID_bin	1.251	1.466	0.854	0.394	-1.626	4.129	747	ft_rep_post
factor(party_numeric)7:PID_bin	0.556	1.686	0.330	0.742	-2.753	3.866	747	ft_rep_post
condition_dummy:social_closeness	-1.073	0.604	-1.776	0.076	-2.259	0.113	747	ft_rep_post
condition_dummy:ft_rep_pre	-0.024	0.040	-0.592	0.554	-0.103	0.055	747	ft_rep_post
condition_dummy:ft_dem_pre	0.007	0.028	0.251	0.802	-0.048	0.062	747	ft_rep_post
condition_dummy:factor(party_numeric)2:PID_bin	0.851	2.049	0.415	0.678	-3.171	4.873	747	ft_rep_post
condition_dummy:factor(party_numeric)3:PID_bin	-5.459	3.720	-1.468	0.143	-12.762	1.843	747	ft_rep_post
condition_dummy:factor(party_numeric)5:PID_bin	-4.709	8.241	-0.571	0.568	-20.887	11.469	747	ft_rep_post
condition_dummy:factor(party_numeric)6:PID_bin	2.554	2.457	1.039	0.299	-2.270	7.378	747	ft_rep_post
condition_dummy:factor(party_numeric)7:PID_bin	0.504	3.194	0.158	0.875	-5.765	6.774	747	ft_rep_post

Hypothesis 6

H6_cov <- lm_robust(ft_dem_post ~ condition_dummy * factor(party_numeric) * PID_bin
                    + condition_dummy * social_closeness
                    + condition_dummy * ft_rep_pre
                    + condition_dummy * ft_dem_pre
                    + ideology_selfreport
                    + maga,
                    data = data_clean, se_type = "HC2")

term	estimate	std.error	statistic	p.value	conf.low	conf.high	df	outcome
(Intercept)	1.479	1.659	0.892	0.373	-1.777	4.735	747	ft_dem_post
condition_dummy	3.192	3.044	1.048	0.295	-2.784	9.168	747	ft_dem_post
factor(party_numeric)2	-3.839	2.199	-1.746	0.081	-8.157	0.478	747	ft_dem_post
factor(party_numeric)3	0.843	1.540	0.548	0.584	-2.180	3.866	747	ft_dem_post
factor(party_numeric)5	-0.880	1.992	-0.442	0.659	-4.790	3.030	747	ft_dem_post
factor(party_numeric)6	-1.134	1.893	-0.599	0.550	-4.851	2.584	747	ft_dem_post
factor(party_numeric)7	-1.432	2.360	-0.607	0.544	-6.066	3.201	747	ft_dem_post
PID_bin	0.795	0.857	0.927	0.354	-0.888	2.477	747	ft_dem_post
social_closeness	0.454	0.555	0.818	0.414	-0.635	1.542	747	ft_dem_post
ft_rep_pre	0.012	0.016	0.774	0.439	-0.019	0.043	747	ft_dem_post
ft_dem_pre	0.957	0.019	49.977	0.000	0.920	0.995	747	ft_dem_post
ideology_selfreport	-0.226	0.324	-0.698	0.485	-0.863	0.410	747	ft_dem_post
maga	-0.018	1.072	-0.017	0.986	-2.123	2.087	747	ft_dem_post
condition_dummy:factor(party_numeric)2	4.748	3.708	1.280	0.201	-2.532	12.029	747	ft_dem_post
condition_dummy:factor(party_numeric)3	-2.409	3.108	-0.775	0.439	-8.511	3.693	747	ft_dem_post
condition_dummy:factor(party_numeric)5	6.483	4.623	1.402	0.161	-2.592	15.559	747	ft_dem_post
condition_dummy:factor(party_numeric)6	-1.327	4.389	-0.302	0.762	-9.944	7.290	747	ft_dem_post
condition_dummy:factor(party_numeric)7	13.409	3.417	3.924	0.000	6.700	20.118	747	ft_dem_post
condition_dummy:PID_bin	0.822	1.685	0.488	0.626	-2.486	4.129	747	ft_dem_post
factor(party_numeric)2:PID_bin	1.914	1.268	1.510	0.132	-0.575	4.403	747	ft_dem_post
factor(party_numeric)3:PID_bin	-2.754	2.118	-1.300	0.194	-6.913	1.405	747	ft_dem_post
factor(party_numeric)5:PID_bin	-1.905	1.178	-1.618	0.106	-4.217	0.406	747	ft_dem_post
factor(party_numeric)6:PID_bin	-0.924	1.715	-0.539	0.590	-4.291	2.442	747	ft_dem_post
factor(party_numeric)7:PID_bin	-0.615	1.370	-0.449	0.654	-3.305	2.075	747	ft_dem_post
condition_dummy:social_closeness	-0.882	0.829	-1.065	0.287	-2.509	0.745	747	ft_dem_post
condition_dummy:ft_rep_pre	-0.026	0.035	-0.740	0.460	-0.093	0.042	747	ft_dem_post
condition_dummy:ft_dem_pre	-0.023	0.036	-0.638	0.524	-0.092	0.047	747	ft_dem_post
condition_dummy:factor(party_numeric)2:PID_bin	-2.960	2.132	-1.388	0.166	-7.146	1.227	747	ft_dem_post
condition_dummy:factor(party_numeric)3:PID_bin	3.636	2.763	1.316	0.189	-1.788	9.059	747	ft_dem_post
condition_dummy:factor(party_numeric)5:PID_bin	-3.076	3.278	-0.938	0.348	-9.510	3.359	747	ft_dem_post
condition_dummy:factor(party_numeric)6:PID_bin	2.695	3.214	0.838	0.402	-3.615	9.006	747	ft_dem_post
condition_dummy:factor(party_numeric)7:PID_bin	-8.109	2.984	-2.718	0.007	-13.967	-2.251	747	ft_dem_post

Scenario-Specific SPV

On the micro level, across each of the six scenarios, I will conduct two Ordinary Least Squares (OLS) linear regression models with heteroskedasticity-robust standard errors, one focused on individual support for political violence and the other on meta-perceptions of Republicans’ support for political violence. For the former, I will create two additional Ordinary Least Squares (OLS) linear regression models with heteroskedasticity-robust standard errors to distinguish between Republicans’ reported SPV and Democrats’ reported SPV within each SPV scenario. Across each of these four models, I plan to regress each of the six scenario-specific measures of SPV separately onto a three-way interaction between condition assignment, the numeric version of the ANES 7-point party identification scale, and the grouped version of partisan self-concept used in Hypotheses 5a through 6c. Additional specified covariates include the interaction between the treatment and social closeness, support for the MAGA movement, and self-reported ideology.

I deviated from the pre-analysis plan by removing the additional specified covariates from each model, as this simplifies the ability to draw conclusions about the average treatment effect. The only covariates included are PID_dummy, which represents the dichotomous effect of being a Democrat versus a Republican, and PID_bin, the grouped version of partisan self-concept.

Scenario: Protesting without a permit

Models

# 1.1: form models for:
# A: Individual SPV (Aggregated)
lm_protest_indiv <- lm_robust(SPV_indiv_protest ~ condition_dummy * PID_dummy * PID_bin,
                              data = data_clean, se_type = "HC2")
# B: Meta-Perceptions of Republican SPV 
lm_protest_meta_R <- lm_robust(SPV_meta_protest ~ condition_dummy * PID_dummy * PID_bin,
                               data = data_clean, se_type = "HC2")
# C: Individual SPV (Reps only)
lm_R_protest <- lm_robust(spv_Rprotest ~ condition_dummy * PID_bin,
                          data = data_clean, se_type = "HC2")
# D: Individual SPV (Dems only)
lm_D_protest <- lm_robust(spv_Dprotest ~ condition_dummy * PID_bin,
                          data = data_clean, se_type = "HC2")

Individual SPV

term	estimate	std.error	statistic	p.value	conf.low	conf.high	df	outcome
(Intercept)	42.795	4.136	10.346	0.000	34.675	50.915	774	SPV_indiv_protest
condition_dummy	15.132	6.411	2.360	0.019	2.547	27.718	774	SPV_indiv_protest
PID_dummy	-26.558	6.080	-4.368	0.000	-38.492	-14.623	774	SPV_indiv_protest
PID_bin	4.368	2.612	1.672	0.095	-0.760	9.496	774	SPV_indiv_protest
condition_dummy:PID_dummy	2.301	10.215	0.225	0.822	-17.752	22.355	774	SPV_indiv_protest
condition_dummy:PID_bin	-5.930	3.960	-1.498	0.135	-13.704	1.843	774	SPV_indiv_protest
PID_dummy:PID_bin	13.091	4.024	3.253	0.001	5.191	20.990	774	SPV_indiv_protest
condition_dummy:PID_dummy:PID_bin	-6.951	6.904	-1.007	0.314	-20.503	6.601	774	SPV_indiv_protest

Meta-Perceptions Republican SPV

term	estimate	std.error	statistic	p.value	conf.low	conf.high	df	outcome
(Intercept)	57.366	4.687	12.238	0.000	48.164	66.567	777	SPV_meta_protest
condition_dummy	10.715	6.962	1.539	0.124	-2.952	24.381	777	SPV_meta_protest
PID_dummy	-9.016	6.890	-1.309	0.191	-22.542	4.510	777	SPV_meta_protest
PID_bin	-0.929	2.888	-0.322	0.748	-6.599	4.741	777	SPV_meta_protest
condition_dummy:PID_dummy	8.520	9.999	0.852	0.394	-11.108	28.148	777	SPV_meta_protest
condition_dummy:PID_bin	-2.945	4.218	-0.698	0.485	-11.225	5.335	777	SPV_meta_protest
PID_dummy:PID_bin	6.206	4.309	1.440	0.150	-2.253	14.665	777	SPV_meta_protest
condition_dummy:PID_dummy:PID_bin	-10.385	6.528	-1.591	0.112	-23.200	2.430	777	SPV_meta_protest

Individual SPV (Republicans Only)

term	estimate	std.error	statistic	p.value	conf.low	conf.high	df	outcome
(Intercept)	16.237	4.455	3.644	0.000	7.463	25.012	253	spv_Rprotest
condition_dummy	17.434	7.953	2.192	0.029	1.772	33.096	253	spv_Rprotest
PID_bin	17.458	3.061	5.703	0.000	11.430	23.487	253	spv_Rprotest
condition_dummy:PID_bin	-12.881	5.655	-2.278	0.024	-24.018	-1.745	253	spv_Rprotest

Individual SPV (Democrats Only)

term	estimate	std.error	statistic	p.value	conf.low	conf.high	df	outcome
(Intercept)	42.795	4.136	10.346	0.000	34.669	50.921	521	spv_Dprotest
condition_dummy	15.132	6.411	2.360	0.019	2.537	27.728	521	spv_Dprotest
PID_bin	4.368	2.612	1.672	0.095	-0.764	9.500	521	spv_Dprotest
condition_dummy:PID_bin	-5.930	3.960	-1.498	0.135	-13.710	1.849	521	spv_Dprotest

Party ID (Dummy)

# form emmeans and contrast matrices for Individual SPV (Aggregated)
emm_protest_indiv <- emmeans(lm_protest_indiv, ~ condition_dummy * PID_dummy,
                             vcov. = vcov(lm_protest_indiv))
emm_protest_indiv_df <- as.data.frame(confint(emm_protest_indiv)) %>%
  mutate(
    condition_dummy = factor(condition_dummy, levels = c(0, 1),
                             labels = c("Control", "Treatment")),
    PID_dummy = factor(PID_dummy, levels = c(0, 1),
                       labels = c("Democrat", "Republican")),
    group = interaction(condition_dummy, PID_dummy))
contr_protest_indiv <- contrast(emm_protest_indiv,
                                method = list(
                                  "Control D vs. Control R"     = c(-1,  0,  1,  0),
                                  "Control D vs. Treatment D"   = c(-1,  1,  0,  0),
                                  "Control R vs. Treatment R"   = c( 0,  0, -1,  1),
                                  "Treatment D vs. Treatment R" = c( 0, -1,  0,  1)))
contr_protest_indiv_df <- as.data.frame(confint(contr_protest_indiv))
contr_table_protest_indiv <- summary(contr_protest_indiv) |>
  as.data.frame() |>
  dplyr::select(contrast, estimate, SE, df, t.ratio, p.value)

contrast	estimate	SE	df	t.ratio	p.value
Control D vs. Control R	-6.830133	2.951703	774	-2.3139637	0.0209308
Control D vs. Treatment D	6.195672	2.550776	774	2.4289363	0.0153699
Control R vs. Treatment R	-1.978148	3.827324	774	-0.5168488	0.6054093
Treatment D vs. Treatment R	-15.003953	3.527367	774	-4.2535843	0.0000236

# form emmeans and contrast matrices for Meta-Perceptions of Republican SPV
emm_protest_meta_R <- emmeans(lm_protest_meta_R, ~ condition_dummy * PID_dummy,
                              vcov. = vcov(lm_protest_meta_R))
emm_protest_meta_R_df <- as.data.frame(confint(emm_protest_meta_R)) %>%
  mutate(
    condition_dummy = factor(condition_dummy, levels = c(0, 1),
                             labels = c("Control", "Treatment")),
    PID_dummy = factor(PID_dummy, levels = c(0, 1),
                       labels = c("Democrat", "Republican")),
    group = interaction(condition_dummy, PID_dummy))
contr_protest_meta_R <- contrast(emm_protest_meta_R,
                                 method = list(
                                   "Control D vs. Control R"     = c(-1,  0,  1,  0),
                                   "Control D vs. Treatment D"   = c(-1,  1,  0,  0),
                                   "Control R vs. Treatment R"   = c( 0,  0, -1,  1),
                                   "Treatment D vs. Treatment R" = c( 0, -1,  0,  1)))
contr_protest_meta_R_df <- as.data.frame(confint(contr_protest_meta_R))
contr_table_protest_meta_R <- summary(contr_protest_meta_R) |>
  as.data.frame() |>
  dplyr::select(contrast, estimate, SE, df, t.ratio, p.value)

contrast	estimate	SE	df	t.ratio	p.value
Control D vs. Control R	0.3363275	2.859830	777	0.1176040	0.9064118
Control D vs. Treatment D	6.2770729	2.672940	777	2.3483778	0.0191051
Control R vs. Treatment R	-0.8533573	3.460719	777	-0.2465838	0.8052955
Treatment D vs. Treatment R	-6.7941027	3.307953	777	-2.0538690	0.0403225

Party ID (7-pt)

# form emmeans and contrast matrices for Individual SPV (Aggregated)
protest_indiv_partyaffil <- lm_robust(SPV_indiv_protest ~ condition_dummy * party_numeric, data = data_clean, se_type = "HC2")
emm_protest_indiv_partyaffil<- emmeans(protest_indiv_partyaffil, ~ condition_dummy * party_numeric,
                                       at = list(party_numeric = c(1, 2, 3, 5, 6, 7)),
                                       vcov. = vcov(protest_indiv_partyaffil))
contr_protest_indiv_partyaffil<- contrast(emm_protest_indiv_partyaffil,
                                          method = list("Treatment - Control" = c(-1, 1)),
                                          by = "party_numeric")
contr_protest_indiv_partyaffil_df <- confint(contr_protest_indiv_partyaffil) |>
  as.data.frame() |>
  mutate(
    party_label = factor(party_numeric, levels = c(1, 2, 3, 5, 6, 7),
                         labels = c("Strong D","Weak D","Lean D","Lean R","Weak R","Strong R")),
    model = "Individual Ratings")

contrast	party_numeric	estimate	SE	df	lower.CL	upper.CL	party_label	model
Treatment - Control	1	7.0566203	2.959668	782	1.2467864	12.866454	Strong D	Individual Ratings
Treatment - Control	2	5.4523769	2.371865	782	0.7964009	10.108353	Weak D	Individual Ratings
Treatment - Control	3	3.8481336	2.128025	782	-0.3291845	8.025452	Lean D	Individual Ratings
Treatment - Control	5	0.6396468	2.905838	782	-5.0645201	6.343814	Lean R	Individual Ratings
Treatment - Control	6	-0.9645965	3.668318	782	-8.1655119	6.236319	Weak R	Individual Ratings
Treatment - Control	7	-2.5688399	4.528338	782	-11.4579778	6.320298	Strong R	Individual Ratings

# form emmeans and contrast matrices for Meta-Perceptions of Republican SPV
protest_meta_R_partyaffil <- lm_robust(SPV_meta_protest ~ condition_dummy * party_numeric, data = data_clean, se_type = "HC2")
emm_protest_meta_R_partyaffil<- emmeans(protest_meta_R_partyaffil, ~ condition_dummy * party_numeric,
                                        at = list(party_numeric = c(1, 2, 3, 5, 6, 7)),
                                        vcov. = vcov(protest_meta_R_partyaffil))
contr_protest_meta_R_partyaffil<- contrast(emm_protest_meta_R_partyaffil,
                                           method = list("Treatment - Control" = c(-1, 1)),
                                           by = "party_numeric")
contr_protest_meta_R_partyaffil_df <- confint(contr_protest_meta_R_partyaffil) |>
  as.data.frame() |>
  mutate(
    party_label = factor(party_numeric, levels = c(1, 2, 3, 5, 6, 7),
                         labels = c("Strong D","Weak D","Lean D","Lean R","Weak R","Strong R")),
    model = "Meta-Perceptions of Republicans' Ratings")

contrast	party_numeric	estimate	SE	df	lower.CL	upper.CL	party_label	model
Treatment - Control	1	6.0762027	3.053646	785	0.0819242	12.070481	Strong D	Meta-Perceptions of Republicans’ Ratings
Treatment - Control	2	5.1819775	2.460482	785	0.3520735	10.011881	Weak D	Meta-Perceptions of Republicans’ Ratings
Treatment - Control	3	4.2877522	2.125921	785	0.1145901	8.460914	Lean D	Meta-Perceptions of Republicans’ Ratings
Treatment - Control	5	2.4993018	2.580640	785	-2.5664697	7.565073	Lean R	Meta-Perceptions of Republicans’ Ratings
Treatment - Control	6	1.6050766	3.214701	785	-4.7053504	7.915503	Weak R	Meta-Perceptions of Republicans’ Ratings
Treatment - Control	7	0.7108514	3.968028	785	-7.0783494	8.500052	Strong R	Meta-Perceptions of Republicans’ Ratings

Party ID (7-pt) + Partisan Self-Concept

lm_protest_indiv_R_7 <- lm_robust(SPV_indiv_protest ~ condition_dummy * party_numeric * PID_bin,
                                  data = data_clean, se_type = "HC2")
emm_protest_indiv_R_7<- emmeans(lm_protest_indiv_R_7, ~ condition_dummy | party_numeric,
                                at = list(party_numeric = c(1, 2, 3, 5, 6, 7)),
                                vcov. = vcov(lm_protest_indiv_R_7))
contr_protest_indiv_R_7<- contrast(emm_protest_indiv_R_7,
                                   method = list("Treatment - Control" = c(-1, 1)),
                                   by = "party_numeric")
contr_protest_indiv_R_7_df <- as.data.frame(confint(contr_protest_indiv_R_7))
contr_table_protest_indiv_R_7_df <- summary(contr_protest_indiv_R_7) |>
  as.data.frame() |>
  dplyr::select(contrast, party_numeric, estimate, SE, df, t.ratio, p.value) |>
  mutate(party_label = factor(party_numeric, levels = c(1, 2, 3, 5, 6, 7),
                              labels = c("Strong D","Weak D","Lean D","Lean R","Weak R","Strong R")))

contrast	party_numeric	estimate	SE	df	t.ratio	p.value	party_label
Treatment - Control	1	6.9789524	3.059601	774	2.2810007	0.0228193	Strong D
Treatment - Control	2	5.4979017	2.452873	774	2.2414131	0.0252820	Weak D
Treatment - Control	3	4.0168511	2.163984	774	1.8562300	0.0638006	Lean D
Treatment - Control	5	1.0547498	2.836854	774	0.3718027	0.7101415	Lean R
Treatment - Control	6	-0.4263008	3.570212	774	-0.1194049	0.9049856	Weak R
Treatment - Control	7	-1.9073515	4.410884	774	-0.4324193	0.6655571	Strong R

lm_protest_meta_R_7 <- lm_robust(SPV_meta_protest ~ condition_dummy * party_numeric * PID_bin,
                                 data = data_clean, se_type = "HC2")
emm_protest_meta_R_7<- emmeans(lm_protest_meta_R_7, ~ condition_dummy | party_numeric,
                               at = list(party_numeric = c(1, 2, 3, 5, 6, 7)),
                               vcov. = vcov(lm_protest_meta_R_7))
contr_protest_meta_R_7<- contrast(emm_protest_meta_R_7,
                                  method = list("Treatment - Control" = c(-1, 1)),
                                  by = "party_numeric")
contr_protest_meta_R_7_df <- as.data.frame(confint(contr_protest_meta_R_7))
contr_table_protest_meta_R_7_df <- summary(contr_protest_meta_R_7) |>
  as.data.frame() |>
  dplyr::select(contrast, party_numeric, estimate, SE, df, t.ratio, p.value) |>
  mutate(party_label = factor(party_numeric, levels = c(1, 2, 3, 5, 6, 7),
                              labels = c("Strong D","Weak D","Lean D","Lean R","Weak R","Strong R")))

contrast	party_numeric	estimate	SE	df	t.ratio	p.value	party_label
Treatment - Control	1	4.970842	3.261950	777	1.5238864	0.1279440	Strong D
Treatment - Control	2	4.407223	2.605459	777	1.6915342	0.0911358	Weak D
Treatment - Control	3	3.843604	2.199456	777	1.7475253	0.0809413	Lean D
Treatment - Control	5	2.716367	2.576675	777	1.0542141	0.2921126	Lean R
Treatment - Control	6	2.152748	3.223618	777	0.6678049	0.5044564	Weak R
Treatment - Control	7	1.589129	4.005654	777	0.3967215	0.6916818	Strong R

Factor models for panel plots

protest_indiv_factor <- lm_robust(SPV_indiv_protest ~ condition_dummy * factor(party_numeric),
                                  data = data_clean, se_type = "HC2")
emm_protest_indiv_factor <- emmeans(protest_indiv_factor, ~ condition_dummy | party_numeric,
                                    vcov. = vcov(protest_indiv_factor))
contr_protest_indiv_factor_df <- confint(contrast(emm_protest_indiv_factor,
                                                  method = list("Treatment - Control" = c(-1, 1)),
                                                  by = "party_numeric")) |>
  as.data.frame() |>
  mutate(party_numeric = as.numeric(as.character(party_numeric)))

protest_meta_R_factor <- lm_robust(SPV_meta_protest ~ condition_dummy * factor(party_numeric),
                                   data = data_clean, se_type = "HC2")
emm_protest_meta_R_factor <- emmeans(protest_meta_R_factor, ~ condition_dummy | party_numeric,
                                     vcov. = vcov(protest_meta_R_factor))
contr_protest_meta_R_factor_df <- confint(contrast(emm_protest_meta_R_factor,
                                                   method = list("Treatment - Control" = c(-1, 1)),
                                                   by = "party_numeric")) |>
  as.data.frame() |>
  mutate(party_numeric = as.numeric(as.character(party_numeric)))

p_protest_indiv <- make_SPV_panels(
  "SPV_indiv_protest",
  contr_protest_indiv_partyaffil_df,
  contr_protest_indiv_factor_df,
  "Individual SPV: Protest\n(raw mean)",
  "Figure: Protest Without a Permit — Individual Ratings",
  color_A = c("Control R" = "firebrick", "Treatment R" = "salmon",
              "Control D" = "hotpink",   "Treatment D" = "pink"))

p_protest_meta_R <- make_SPV_panels(
  "SPV_meta_protest",
  contr_protest_meta_R_partyaffil_df,
  contr_protest_meta_R_factor_df,
  "Meta-Perceptions of Republican SPV: Protest\n(raw mean)",
  "Figure: Protest Without a Permit — Meta-Perceptions",
  color_A = c("Control R" = "firebrick", "Treatment R" = "salmon",
              "Control D" = "hotpink",   "Treatment D" = "pink"))

Scenario: Vandalizing opposing party’s signs

Models

# 1.1: form models for:
# A: Individual SPV (Aggregated)
lm_vandalize_indiv <- lm_robust(SPV_indiv_vandalize ~ condition_dummy * PID_dummy * PID_bin,
                              data = data_clean, se_type = "HC2")
# B: Meta-Perceptions of Republican SPV 
lm_vandalize_meta_R <- lm_robust(SPV_meta_vandalize ~ condition_dummy * PID_dummy * PID_bin,
                               data = data_clean, se_type = "HC2")
# C: Individual SPV (Reps only)
lm_R_vandalize <- lm_robust(spv_Rvandalize ~ condition_dummy * PID_bin,
                          data = data_clean, se_type = "HC2")
# D: Individual SPV (Dems only)
lm_D_vandalize <- lm_robust(spv_Dvandalize ~ condition_dummy * PID_bin,
                          data = data_clean, se_type = "HC2")

Individual SPV

term	estimate	std.error	statistic	p.value	conf.low	conf.high	df	outcome
(Intercept)	32.932	4.171	7.895	0.000	24.744	41.121	772	SPV_indiv_vandalize
condition_dummy	7.143	7.003	1.020	0.308	-6.605	20.891	772	SPV_indiv_vandalize
PID_dummy	-28.169	6.052	-4.655	0.000	-40.049	-16.289	772	SPV_indiv_vandalize
PID_bin	-2.083	2.547	-0.818	0.414	-7.083	2.917	772	SPV_indiv_vandalize
condition_dummy:PID_dummy	2.273	9.608	0.237	0.813	-16.587	21.133	772	SPV_indiv_vandalize
condition_dummy:PID_bin	-0.132	4.183	-0.032	0.975	-8.345	8.080	772	SPV_indiv_vandalize
PID_dummy:PID_bin	15.264	3.983	3.832	0.000	7.445	23.082	772	SPV_indiv_vandalize
condition_dummy:PID_dummy:PID_bin	-6.212	6.333	-0.981	0.327	-18.644	6.220	772	SPV_indiv_vandalize

Meta-Perceptions Republican SPV

term	estimate	std.error	statistic	p.value	conf.low	conf.high	df	outcome
(Intercept)	47.768	4.808	9.934	0.000	38.329	57.207	773	SPV_meta_vandalize
condition_dummy	13.445	7.576	1.775	0.076	-1.427	28.317	773	SPV_meta_vandalize
PID_dummy	-22.082	6.643	-3.324	0.001	-35.122	-9.042	773	SPV_meta_vandalize
PID_bin	-0.377	2.990	-0.126	0.900	-6.247	5.492	773	SPV_meta_vandalize
condition_dummy:PID_dummy	11.520	11.305	1.019	0.308	-10.671	33.712	773	SPV_meta_vandalize
condition_dummy:PID_bin	-4.542	4.563	-0.995	0.320	-13.500	4.415	773	SPV_meta_vandalize
PID_dummy:PID_bin	8.770	4.349	2.017	0.044	0.233	17.307	773	SPV_meta_vandalize
condition_dummy:PID_dummy:PID_bin	-10.516	7.301	-1.440	0.150	-24.847	3.815	773	SPV_meta_vandalize

Individual SPV (Republicans Only)

term	estimate	std.error	statistic	p.value	conf.low	conf.high	df	outcome
(Intercept)	4.763	4.385	1.086	0.278	-3.872	13.398	254	spv_Rvandalize
condition_dummy	9.416	6.577	1.432	0.153	-3.537	22.369	254	spv_Rvandalize
PID_bin	13.181	3.062	4.305	0.000	7.151	19.211	254	spv_Rvandalize
condition_dummy:PID_bin	-6.344	4.754	-1.334	0.183	-15.707	3.019	254	spv_Rvandalize

Individual SPV (Democrats Only)

term	estimate	std.error	statistic	p.value	conf.low	conf.high	df	outcome
(Intercept)	32.932	4.171	7.895	0.000	24.738	41.127	518	spv_Dvandalize
condition_dummy	7.143	7.003	1.020	0.308	-6.616	20.902	518	spv_Dvandalize
PID_bin	-2.083	2.547	-0.818	0.414	-7.087	2.921	518	spv_Dvandalize
condition_dummy:PID_bin	-0.132	4.183	-0.032	0.975	-8.351	8.086	518	spv_Dvandalize

Party ID (Dummy)

# form emmeans and contrast matrices for Individual SPV (Aggregated)
emm_vandalize_indiv <- emmeans(lm_vandalize_indiv, ~ condition_dummy * PID_dummy,
                             vcov. = vcov(lm_vandalize_indiv))
emm_vandalize_indiv_df <- as.data.frame(confint(emm_vandalize_indiv)) %>%
  mutate(
    condition_dummy = factor(condition_dummy, levels = c(0, 1),
                             labels = c("Control", "Treatment")),
    PID_dummy = factor(PID_dummy, levels = c(0, 1),
                       labels = c("Democrat", "Republican")),
    group = interaction(condition_dummy, PID_dummy))
contr_vandalize_indiv <- contrast(emm_vandalize_indiv,
                                method = list(
                                  "Control D vs. Control R"     = c(-1,  0,  1,  0),
                                  "Control D vs. Treatment D"   = c(-1,  1,  0,  0),
                                  "Control R vs. Treatment R"   = c( 0,  0, -1,  1),
                                  "Treatment D vs. Treatment R" = c( 0, -1,  0,  1)))
contr_vandalize_indiv_df <- as.data.frame(confint(contr_vandalize_indiv))
contr_table_vandalize_indiv <- summary(contr_vandalize_indiv) |>
  as.data.frame() |>
  dplyr::select(contrast, estimate, SE, df, t.ratio, p.value)

contrast	estimate	SE	df	t.ratio	p.value
Control D vs. Control R	-5.1666652	2.753057	772	-1.8767011	0.0609360
Control D vs. Treatment D	6.9436643	2.563970	772	2.7081689	0.0069153
Control R vs. Treatment R	-0.1446896	3.389855	772	-0.0426831	0.9659652
Treatment D vs. Treatment R	-12.2550191	3.238169	772	-3.7845525	0.0001659

# form emmeans and contrast matrices for Meta-Perceptions of Republican SPV
emm_vandalize_meta_R <- emmeans(lm_vandalize_meta_R, ~ condition_dummy * PID_dummy,
                              vcov. = vcov(lm_vandalize_meta_R))
emm_vandalize_meta_R_df <- as.data.frame(confint(emm_vandalize_meta_R)) %>%
  mutate(
    condition_dummy = factor(condition_dummy, levels = c(0, 1),
                             labels = c("Control", "Treatment")),
    PID_dummy = factor(PID_dummy, levels = c(0, 1),
                       labels = c("Democrat", "Republican")),
    group = interaction(condition_dummy, PID_dummy))
contr_vandalize_meta_R <- contrast(emm_vandalize_meta_R,
                                 method = list(
                                   "Control D vs. Control R"     = c(-1,  0,  1,  0),
                                   "Control D vs. Treatment D"   = c(-1,  1,  0,  0),
                                   "Control R vs. Treatment R"   = c( 0,  0, -1,  1),
                                   "Treatment D vs. Treatment R" = c( 0, -1,  0,  1)))
contr_vandalize_meta_R_df <- as.data.frame(confint(contr_vandalize_meta_R))
contr_table_vandalize_meta_R <- summary(contr_vandalize_meta_R) |>
  as.data.frame() |>
  dplyr::select(contrast, estimate, SE, df, t.ratio, p.value)

contrast	estimate	SE	df	t.ratio	p.value
Control D vs. Control R	-8.865556	2.946239	773	-3.0091100	0.0027053
Control D vs. Treatment D	6.599702	2.892820	773	2.2814076	0.0227955
Control R vs. Treatment R	2.271952	3.516133	773	0.6461508	0.5183735
Treatment D vs. Treatment R	-13.193306	3.471495	773	-3.8004682	0.0001558

Party ID (7-pt)

# form emmeans and contrast matrices for Individual SPV (Aggregated)
vandalize_indiv_partyaffil <- lm_robust(SPV_indiv_vandalize ~ condition_dummy * party_numeric, data = data_clean, se_type = "HC2")
emm_vandalize_indiv_partyaffil<- emmeans(vandalize_indiv_partyaffil, ~ condition_dummy * party_numeric,
                                       at = list(party_numeric = c(1, 2, 3, 5, 6, 7)),
                                       vcov. = vcov(vandalize_indiv_partyaffil))
contr_vandalize_indiv_partyaffil<- contrast(emm_vandalize_indiv_partyaffil,
                                          method = list("Treatment - Control" = c(-1, 1)),
                                          by = "party_numeric")
contr_vandalize_indiv_partyaffil_df <- confint(contr_vandalize_indiv_partyaffil) |>
  as.data.frame() |>
  mutate(
    party_label = factor(party_numeric, levels = c(1, 2, 3, 5, 6, 7),
                         labels = c("Strong D","Weak D","Lean D","Lean R","Weak R","Strong R")),
    model = "Individual Ratings")

contrast	party_numeric	estimate	SE	df	lower.CL	upper.CL	party_label	model
Treatment - Control	1	8.0913299	2.930071	780	2.3395716	13.843088	Strong D	Individual Ratings
Treatment - Control	2	6.4550897	2.355784	780	1.8306622	11.079517	Weak D	Individual Ratings
Treatment - Control	3	4.8188496	2.052619	780	0.7895374	8.848162	Lean D	Individual Ratings
Treatment - Control	5	1.5463693	2.576369	780	-3.5110681	6.603807	Lean R	Individual Ratings
Treatment - Control	6	-0.0898709	3.224682	780	-6.4199547	6.240213	Weak R	Individual Ratings
Treatment - Control	7	-1.7261110	3.982286	780	-9.5433785	6.091156	Strong R	Individual Ratings

# form emmeans and contrast matrices for Meta-Perceptions of Republican SPV
vandalize_meta_R_partyaffil <- lm_robust(SPV_meta_vandalize ~ condition_dummy * party_numeric, data = data_clean, se_type = "HC2")
emm_vandalize_meta_R_partyaffil<- emmeans(vandalize_meta_R_partyaffil, ~ condition_dummy * party_numeric,
                                        at = list(party_numeric = c(1, 2, 3, 5, 6, 7)),
                                        vcov. = vcov(vandalize_meta_R_partyaffil))
contr_vandalize_meta_R_partyaffil<- contrast(emm_vandalize_meta_R_partyaffil,
                                           method = list("Treatment - Control" = c(-1, 1)),
                                           by = "party_numeric")
contr_vandalize_meta_R_partyaffil_df <- confint(contr_vandalize_meta_R_partyaffil) |>
  as.data.frame() |>
  mutate(
    party_label = factor(party_numeric, levels = c(1, 2, 3, 5, 6, 7),
                         labels = c("Strong D","Weak D","Lean D","Lean R","Weak R","Strong R")),
    model = "Meta-Perceptions of Republicans' Ratings")

contrast	party_numeric	estimate	SE	df	lower.CL	upper.CL	party_label	model
Treatment - Control	1	6.532577	3.269176	781	0.1151645	12.949988	Strong D	Meta-Perceptions of Republicans’ Ratings
Treatment - Control	2	6.034259	2.641108	781	0.8497477	11.218770	Weak D	Meta-Perceptions of Republicans’ Ratings
Treatment - Control	3	5.535941	2.270913	781	1.0781248	9.993758	Lean D	Meta-Perceptions of Republicans’ Ratings
Treatment - Control	5	4.539306	2.683505	781	-0.7284307	9.807042	Lean R	Meta-Perceptions of Republicans’ Ratings
Treatment - Control	6	4.040988	3.326223	781	-2.4884067	10.570383	Weak R	Meta-Perceptions of Republicans’ Ratings
Treatment - Control	7	3.542671	4.101215	781	-4.5080400	11.593381	Strong R	Meta-Perceptions of Republicans’ Ratings

Party ID (7-pt) + Partisan Self-Concept

lm_vandalize_indiv_R_7 <- lm_robust(SPV_indiv_vandalize ~ condition_dummy * party_numeric * PID_bin,
                                  data = data_clean, se_type = "HC2")
emm_vandalize_indiv_R_7<- emmeans(lm_vandalize_indiv_R_7, ~ condition_dummy | party_numeric,
                                at = list(party_numeric = c(1, 2, 3, 5, 6, 7)),
                                vcov. = vcov(lm_vandalize_indiv_R_7))
contr_vandalize_indiv_R_7<- contrast(emm_vandalize_indiv_R_7,
                                   method = list("Treatment - Control" = c(-1, 1)),
                                   by = "party_numeric")
contr_vandalize_indiv_R_7_df <- as.data.frame(confint(contr_vandalize_indiv_R_7))
contr_table_vandalize_indiv_R_7_df <- summary(contr_vandalize_indiv_R_7) |>
  as.data.frame() |>
  dplyr::select(contrast, party_numeric, estimate, SE, df, t.ratio, p.value) |>
  mutate(party_label = factor(party_numeric, levels = c(1, 2, 3, 5, 6, 7),
                              labels = c("Strong D","Weak D","Lean D","Lean R","Weak R","Strong R")))

contrast	party_numeric	estimate	SE	df	t.ratio	p.value	party_label
Treatment - Control	1	7.5900529	3.192076	772	2.3777796	0.0176598	Strong D
Treatment - Control	2	6.2708727	2.539107	772	2.4697156	0.0137373	Weak D
Treatment - Control	3	4.9516925	2.132891	772	2.3215870	0.0205146	Lean D
Treatment - Control	5	2.3133321	2.507363	772	0.9226156	0.3564958	Lean R
Treatment - Control	6	0.9941519	3.149977	772	0.3156061	0.7523868	Weak R
Treatment - Control	7	-0.3250284	3.924574	772	-0.0828188	0.9340171	Strong R

lm_vandalize_meta_R_7 <- lm_robust(SPV_meta_vandalize ~ condition_dummy * party_numeric * PID_bin,
                                 data = data_clean, se_type = "HC2")
emm_vandalize_meta_R_7<- emmeans(lm_vandalize_meta_R_7, ~ condition_dummy | party_numeric,
                               at = list(party_numeric = c(1, 2, 3, 5, 6, 7)),
                               vcov. = vcov(lm_vandalize_meta_R_7))
contr_vandalize_meta_R_7<- contrast(emm_vandalize_meta_R_7,
                                  method = list("Treatment - Control" = c(-1, 1)),
                                  by = "party_numeric")
contr_vandalize_meta_R_7_df <- as.data.frame(confint(contr_vandalize_meta_R_7))
contr_table_vandalize_meta_R_7_df <- summary(contr_vandalize_meta_R_7) |>
  as.data.frame() |>
  dplyr::select(contrast, party_numeric, estimate, SE, df, t.ratio, p.value) |>
  mutate(party_label = factor(party_numeric, levels = c(1, 2, 3, 5, 6, 7),
                              labels = c("Strong D","Weak D","Lean D","Lean R","Weak R","Strong R")))

contrast	party_numeric	estimate	SE	df	t.ratio	p.value	party_label
Treatment - Control	1	5.531969	3.531539	773	1.566447	0.1176532	Strong D
Treatment - Control	2	5.320441	2.817176	773	1.888572	0.0593233	Weak D
Treatment - Control	3	5.108912	2.349322	773	2.174632	0.0299602	Lean D
Treatment - Control	5	4.685854	2.653172	773	1.766133	0.0777683	Lean R
Treatment - Control	6	4.474325	3.313080	773	1.350503	0.1772499	Weak R
Treatment - Control	7	4.262796	4.126959	773	1.032914	0.3019669	Strong R

Factor models for panel plots

vandalize_indiv_factor <- lm_robust(SPV_indiv_vandalize ~ condition_dummy * factor(party_numeric),
                                  data = data_clean, se_type = "HC2")
emm_vandalize_indiv_factor <- emmeans(vandalize_indiv_factor, ~ condition_dummy | party_numeric,
                                    vcov. = vcov(vandalize_indiv_factor))
contr_vandalize_indiv_factor_df <- confint(contrast(emm_vandalize_indiv_factor,
                                                  method = list("Treatment - Control" = c(-1, 1)),
                                                  by = "party_numeric")) |>
  as.data.frame() |>
  mutate(party_numeric = as.numeric(as.character(party_numeric)))

vandalize_meta_R_factor <- lm_robust(SPV_meta_vandalize ~ condition_dummy * factor(party_numeric),
                                   data = data_clean, se_type = "HC2")
emm_vandalize_meta_R_factor <- emmeans(vandalize_meta_R_factor, ~ condition_dummy | party_numeric,
                                     vcov. = vcov(vandalize_meta_R_factor))
contr_vandalize_meta_R_factor_df <- confint(contrast(emm_vandalize_meta_R_factor,
                                                   method = list("Treatment - Control" = c(-1, 1)),
                                                   by = "party_numeric")) |>
  as.data.frame() |>
  mutate(party_numeric = as.numeric(as.character(party_numeric)))

p_vandalize_indiv <- make_SPV_panels(
  "SPV_indiv_vandalize",
  contr_vandalize_indiv_partyaffil_df,
  contr_vandalize_indiv_factor_df,
  "Individual SPV: Vandalize\n(raw mean)",
  "Figure: Vandalizing Opposing Party's Signs — Individual Ratings",
  color_A = c("Control R" = "darkorange",  "Treatment R" = "lightsalmon",
              "Control D" = "coral",        "Treatment D" = "peachpuff"))

p_vandalize_meta_R <- make_SPV_panels(
  "SPV_meta_vandalize",
  contr_vandalize_meta_R_partyaffil_df,
  contr_vandalize_meta_R_factor_df,
  "Meta-Perceptions of Republican SPV: Vandalize\n(raw mean)",
  "Figure: Vandalizing Opposing Party's Signs — Meta-Perceptions",
  color_A = c("Control R" = "darkorange",  "Treatment R" = "lightsalmon",
              "Control D" = "coral",        "Treatment D" = "peachpuff"))

Scenario: Messaging threats online to an out-party representative

Models

# 1.1: form models for:
# A: Individual SPV (Aggregated)
lm_message_indiv <- lm_robust(SPV_indiv_message ~ condition_dummy * PID_dummy * PID_bin,
                              data = data_clean, se_type = "HC2")
# B: Meta-Perceptions of Republican SPV 
lm_message_meta_R <- lm_robust(SPV_meta_message ~ condition_dummy * PID_dummy * PID_bin,
                               data = data_clean, se_type = "HC2")
# C: Individual SPV (Reps only)
lm_R_message <- lm_robust(spv_Rmessage ~ condition_dummy * PID_bin,
                          data = data_clean, se_type = "HC2")
# D: Individual SPV (Dems only)
lm_D_message <- lm_robust(spv_Dmessage ~ condition_dummy * PID_bin,
                          data = data_clean, se_type = "HC2")

Individual SPV

term	estimate	std.error	statistic	p.value	conf.low	conf.high	df	outcome
(Intercept)	11.479	3.016	3.806	0.000	5.559	17.399	776	SPV_indiv_message
condition_dummy	9.043	5.074	1.782	0.075	-0.917	19.004	776	SPV_indiv_message
PID_dummy	-10.487	5.260	-1.994	0.047	-20.813	-0.161	776	SPV_indiv_message
PID_bin	1.658	1.939	0.855	0.393	-2.149	5.464	776	SPV_indiv_message
condition_dummy:PID_dummy	-10.092	7.429	-1.358	0.175	-24.675	4.492	776	SPV_indiv_message
condition_dummy:PID_bin	-4.948	3.103	-1.595	0.111	-11.039	1.142	776	SPV_indiv_message
PID_dummy:PID_bin	8.394	3.593	2.336	0.020	1.340	15.448	776	SPV_indiv_message
condition_dummy:PID_dummy:PID_bin	4.245	5.458	0.778	0.437	-6.470	14.960	776	SPV_indiv_message

Meta-Perceptions Republican SPV

term	estimate	std.error	statistic	p.value	conf.low	conf.high	df	outcome
(Intercept)	32.821	4.818	6.813	0.000	23.364	42.278	775	SPV_meta_message
condition_dummy	16.817	7.474	2.250	0.025	2.144	31.489	775	SPV_meta_message
PID_dummy	-25.232	6.005	-4.202	0.000	-37.021	-13.443	775	SPV_meta_message
PID_bin	0.340	2.940	0.116	0.908	-5.432	6.112	775	SPV_meta_message
condition_dummy:PID_dummy	2.535	10.416	0.243	0.808	-17.911	22.982	775	SPV_meta_message
condition_dummy:PID_bin	-7.061	4.498	-1.570	0.117	-15.890	1.769	775	SPV_meta_message
PID_dummy:PID_bin	9.672	3.970	2.436	0.015	1.880	17.465	775	SPV_meta_message
condition_dummy:PID_dummy:PID_bin	-3.328	6.882	-0.484	0.629	-16.837	10.180	775	SPV_meta_message

Individual SPV (Republicans Only)

term	estimate	std.error	statistic	p.value	conf.low	conf.high	df	outcome
(Intercept)	0.992	4.310	0.230	0.818	-7.496	9.480	254	spv_Rmessage
condition_dummy	-1.048	5.426	-0.193	0.847	-11.735	9.638	254	spv_Rmessage
PID_bin	10.052	3.025	3.323	0.001	4.094	16.009	254	spv_Rmessage
condition_dummy:PID_bin	-0.703	4.491	-0.157	0.876	-9.547	8.140	254	spv_Rmessage

Individual SPV (Democrats Only)

term	estimate	std.error	statistic	p.value	conf.low	conf.high	df	outcome
(Intercept)	11.479	3.016	3.806	0.000	5.555	17.403	522	spv_Dmessage
condition_dummy	9.043	5.074	1.782	0.075	-0.924	19.011	522	spv_Dmessage
PID_bin	1.658	1.939	0.855	0.393	-2.152	5.467	522	spv_Dmessage
condition_dummy:PID_bin	-4.948	3.103	-1.595	0.111	-11.044	1.147	522	spv_Dmessage

Party ID (Dummy)

# form emmeans and contrast matrices for Individual SPV (Aggregated)
emm_message_indiv <- emmeans(lm_message_indiv, ~ condition_dummy * PID_dummy,
                             vcov. = vcov(lm_message_indiv))
emm_message_indiv_df <- as.data.frame(confint(emm_message_indiv)) %>%
  mutate(
    condition_dummy = factor(condition_dummy, levels = c(0, 1),
                             labels = c("Control", "Treatment")),
    PID_dummy = factor(PID_dummy, levels = c(0, 1),
                       labels = c("Democrat", "Republican")),
    group = interaction(condition_dummy, PID_dummy))
contr_message_indiv <- contrast(emm_message_indiv,
                                method = list(
                                  "Control D vs. Control R"     = c(-1,  0,  1,  0),
                                  "Control D vs. Treatment D"   = c(-1,  1,  0,  0),
                                  "Control R vs. Treatment R"   = c( 0,  0, -1,  1),
                                  "Treatment D vs. Treatment R" = c( 0, -1,  0,  1)))
contr_message_indiv_df <- as.data.frame(confint(contr_message_indiv))
contr_table_message_indiv <- summary(contr_message_indiv) |>
  as.data.frame() |>
  dplyr::select(contrast, estimate, SE, df, t.ratio, p.value)

contrast	estimate	SE	df	t.ratio	p.value
Control D vs. Control R	2.162887	2.665734	776	0.8113664	0.4174041
Control D vs. Treatment D	1.586123	2.048571	776	0.7742580	0.4390141
Control R vs. Treatment R	-2.108281	3.401649	776	-0.6197822	0.5355832
Treatment D vs. Treatment R	-1.531517	2.943080	776	-0.5203789	0.6029480

# form emmeans and contrast matrices for Meta-Perceptions of Republican SPV
emm_message_meta_R <- emmeans(lm_message_meta_R, ~ condition_dummy * PID_dummy,
                              vcov. = vcov(lm_message_meta_R))
emm_message_meta_R_df <- as.data.frame(confint(emm_message_meta_R)) %>%
  mutate(
    condition_dummy = factor(condition_dummy, levels = c(0, 1),
                             labels = c("Control", "Treatment")),
    PID_dummy = factor(PID_dummy, levels = c(0, 1),
                       labels = c("Democrat", "Republican")),
    group = interaction(condition_dummy, PID_dummy))
contr_message_meta_R <- contrast(emm_message_meta_R,
                                 method = list(
                                   "Control D vs. Control R"     = c(-1,  0,  1,  0),
                                   "Control D vs. Treatment D"   = c(-1,  1,  0,  0),
                                   "Control R vs. Treatment R"   = c( 0,  0, -1,  1),
                                   "Treatment D vs. Treatment R" = c( 0, -1,  0,  1)))
contr_message_meta_R_df <- as.data.frame(confint(contr_message_meta_R))
contr_table_message_meta_R <- summary(contr_message_meta_R) |>
  as.data.frame() |>
  dplyr::select(contrast, estimate, SE, df, t.ratio, p.value)

contrast	estimate	SE	df	t.ratio	p.value
Control D vs. Control R	-10.656018	2.905269	775	-3.667825	0.0002613
Control D vs. Treatment D	6.176294	2.905850	775	2.125469	0.0338632
Control R vs. Treatment R	3.695746	3.556667	775	1.039104	0.2990806
Treatment D vs. Treatment R	-13.136567	3.557142	775	-3.693012	0.0002372

Party ID (7-pt)

# form emmeans and contrast matrices for Individual SPV (Aggregated)
message_indiv_partyaffil <- lm_robust(SPV_indiv_message ~ condition_dummy * party_numeric, data = data_clean, se_type = "HC2")
emm_message_indiv_partyaffil<- emmeans(message_indiv_partyaffil, ~ condition_dummy * party_numeric,
                                       at = list(party_numeric = c(1, 2, 3, 5, 6, 7)),
                                       vcov. = vcov(message_indiv_partyaffil))
contr_message_indiv_partyaffil<- contrast(emm_message_indiv_partyaffil,
                                          method = list("Treatment - Control" = c(-1, 1)),
                                          by = "party_numeric")
contr_message_indiv_partyaffil_df <- confint(contr_message_indiv_partyaffil) |>
  as.data.frame() |>
  mutate(
    party_label = factor(party_numeric, levels = c(1, 2, 3, 5, 6, 7),
                         labels = c("Strong D","Weak D","Lean D","Lean R","Weak R","Strong R")),
    model = "Individual Ratings")

contrast	party_numeric	estimate	SE	df	lower.CL	upper.CL	party_label	model
Treatment - Control	1	1.4912600	2.324198	784	-3.071128	6.053648	Strong D	Individual Ratings
Treatment - Control	2	0.7895172	1.870194	784	-2.881664	4.460698	Weak D	Individual Ratings
Treatment - Control	3	0.0877744	1.734743	784	-3.317516	3.493065	Lean D	Individual Ratings
Treatment - Control	5	-1.3157113	2.506126	784	-6.235223	3.603801	Lean R	Individual Ratings
Treatment - Control	6	-2.0174541	3.168643	784	-8.237483	4.202575	Weak R	Individual Ratings
Treatment - Control	7	-2.7191969	3.900757	784	-10.376361	4.937967	Strong R	Individual Ratings

# form emmeans and contrast matrices for Meta-Perceptions of Republican SPV
message_meta_R_partyaffil <- lm_robust(SPV_meta_message ~ condition_dummy * party_numeric, data = data_clean, se_type = "HC2")
emm_message_meta_R_partyaffil<- emmeans(message_meta_R_partyaffil, ~ condition_dummy * party_numeric,
                                        at = list(party_numeric = c(1, 2, 3, 5, 6, 7)),
                                        vcov. = vcov(message_meta_R_partyaffil))
contr_message_meta_R_partyaffil<- contrast(emm_message_meta_R_partyaffil,
                                           method = list("Treatment - Control" = c(-1, 1)),
                                           by = "party_numeric")
contr_message_meta_R_partyaffil_df <- confint(contr_message_meta_R_partyaffil) |>
  as.data.frame() |>
  mutate(
    party_label = factor(party_numeric, levels = c(1, 2, 3, 5, 6, 7),
                         labels = c("Strong D","Weak D","Lean D","Lean R","Weak R","Strong R")),
    model = "Meta-Perceptions of Republicans' Ratings")

contrast	party_numeric	estimate	SE	df	lower.CL	upper.CL	party_label	model
Treatment - Control	1	5.807870	3.289241	783	-0.6489039	12.264644	Strong D	Meta-Perceptions of Republicans’ Ratings
Treatment - Control	2	5.593071	2.658699	783	0.3740494	10.812093	Weak D	Meta-Perceptions of Republicans’ Ratings
Treatment - Control	3	5.378273	2.285402	783	0.8920331	9.864513	Lean D	Meta-Perceptions of Republicans’ Ratings
Treatment - Control	5	4.948676	2.691882	783	-0.3354840	10.232836	Lean R	Meta-Perceptions of Republicans’ Ratings
Treatment - Control	6	4.733878	3.333919	783	-1.8106001	11.278356	Weak R	Meta-Perceptions of Republicans’ Ratings
Treatment - Control	7	4.519079	4.109575	783	-3.5480098	12.586169	Strong R	Meta-Perceptions of Republicans’ Ratings

Party ID (7-pt) + Partisan Self-Concept

lm_message_indiv_R_7 <- lm_robust(SPV_indiv_message ~ condition_dummy * party_numeric * PID_bin,
                                  data = data_clean, se_type = "HC2")
emm_message_indiv_R_7<- emmeans(lm_message_indiv_R_7, ~ condition_dummy | party_numeric,
                                at = list(party_numeric = c(1, 2, 3, 5, 6, 7)),
                                vcov. = vcov(lm_message_indiv_R_7))
contr_message_indiv_R_7<- contrast(emm_message_indiv_R_7,
                                   method = list("Treatment - Control" = c(-1, 1)),
                                   by = "party_numeric")
contr_message_indiv_R_7_df <- as.data.frame(confint(contr_message_indiv_R_7))
contr_table_message_indiv_R_7_df <- summary(contr_message_indiv_R_7) |>
  as.data.frame() |>
  dplyr::select(contrast, party_numeric, estimate, SE, df, t.ratio, p.value) |>
  mutate(party_label = factor(party_numeric, levels = c(1, 2, 3, 5, 6, 7),
                              labels = c("Strong D","Weak D","Lean D","Lean R","Weak R","Strong R")))

contrast	party_numeric	estimate	SE	df	t.ratio	p.value	party_label
Treatment - Control	1	2.4302838	2.392370	776	1.0158477	0.3100185	Strong D
Treatment - Control	2	1.5926309	1.961566	776	0.8119181	0.4170876	Weak D
Treatment - Control	3	0.7549779	1.804752	776	0.4183277	0.6758233	Lean D
Treatment - Control	5	-0.9203281	2.435399	776	-0.3778961	0.7056111	Lean R
Treatment - Control	6	-1.7579810	3.032461	776	-0.5797209	0.5622711	Weak R
Treatment - Control	7	-2.5956340	3.707500	776	-0.7001037	0.4840724	Strong R

lm_message_meta_R_7 <- lm_robust(SPV_meta_message ~ condition_dummy * party_numeric * PID_bin,
                                 data = data_clean, se_type = "HC2")
emm_message_meta_R_7<- emmeans(lm_message_meta_R_7, ~ condition_dummy | party_numeric,
                               at = list(party_numeric = c(1, 2, 3, 5, 6, 7)),
                               vcov. = vcov(lm_message_meta_R_7))
contr_message_meta_R_7<- contrast(emm_message_meta_R_7,
                                  method = list("Treatment - Control" = c(-1, 1)),
                                  by = "party_numeric")
contr_message_meta_R_7_df <- as.data.frame(confint(contr_message_meta_R_7))
contr_table_message_meta_R_7_df <- summary(contr_message_meta_R_7) |>
  as.data.frame() |>
  dplyr::select(contrast, party_numeric, estimate, SE, df, t.ratio, p.value) |>
  mutate(party_label = factor(party_numeric, levels = c(1, 2, 3, 5, 6, 7),
                              labels = c("Strong D","Weak D","Lean D","Lean R","Weak R","Strong R")))

contrast	party_numeric	estimate	SE	df	t.ratio	p.value	party_label
Treatment - Control	1	6.025631	3.479417	775	1.731793	0.0837083	Strong D
Treatment - Control	2	5.814006	2.800035	775	2.076405	0.0381856	Weak D
Treatment - Control	3	5.602382	2.355789	775	2.378134	0.0176420	Lean D
Treatment - Control	5	5.179133	2.626008	775	1.972246	0.0489367	Lean R
Treatment - Control	6	4.967509	3.245389	775	1.530636	0.1262675	Weak R
Treatment - Control	7	4.755884	4.018068	775	1.183625	0.2369246	Strong R

Factor models for panel plots

message_indiv_factor <- lm_robust(SPV_indiv_message ~ condition_dummy * factor(party_numeric),
                                  data = data_clean, se_type = "HC2")
emm_message_indiv_factor <- emmeans(message_indiv_factor, ~ condition_dummy | party_numeric,
                                    vcov. = vcov(message_indiv_factor))
contr_message_indiv_factor_df <- confint(contrast(emm_message_indiv_factor,
                                                  method = list("Treatment - Control" = c(-1, 1)),
                                                  by = "party_numeric")) |>
  as.data.frame() |>
  mutate(party_numeric = as.numeric(as.character(party_numeric)))

message_meta_R_factor <- lm_robust(SPV_meta_message ~ condition_dummy * factor(party_numeric),
                                   data = data_clean, se_type = "HC2")
emm_message_meta_R_factor <- emmeans(message_meta_R_factor, ~ condition_dummy | party_numeric,
                                     vcov. = vcov(message_meta_R_factor))
contr_message_meta_R_factor_df <- confint(contrast(emm_message_meta_R_factor,
                                                   method = list("Treatment - Control" = c(-1, 1)),
                                                   by = "party_numeric")) |>
  as.data.frame() |>
  mutate(party_numeric = as.numeric(as.character(party_numeric)))

p_message_indiv <- make_SPV_panels(
  "SPV_indiv_message",
  contr_message_indiv_partyaffil_df,
  contr_message_indiv_factor_df,
  "Individual SPV: Message\n(raw mean)",
  "Figure: Messaging Threats Online — Individual Ratings",
  color_A = c("Control R" = "goldenrod", "Treatment R" = "khaki",
              "Control D" = "gold",       "Treatment D" = "wheat"))

p_message_meta_R <- make_SPV_panels(
  "SPV_meta_message",
  contr_message_meta_R_partyaffil_df,
  contr_message_meta_R_factor_df,
  "Meta-Perceptions of Republican SPV: Message\n(raw mean)",
  "Figure: Messaging Threats Online — Meta-Perceptions",
  color_A = c("Control R" = "goldenrod", "Treatment R" = "khaki",
              "Control D" = "gold",       "Treatment D" = "wheat"))

Scenario: Physically assaulting an out-party member

Models

# 1.1: form models for:
# A: Individual SPV (Aggregated)
lm_assault_indiv <- lm_robust(SPV_indiv_assault ~ condition_dummy * PID_dummy * PID_bin,
                              data = data_clean, se_type = "HC2")
# B: Meta-Perceptions of Republican SPV 
lm_assault_meta_R <- lm_robust(SPV_meta_assault ~ condition_dummy * PID_dummy * PID_bin,
                               data = data_clean, se_type = "HC2")
# C: Individual SPV (Reps only)
lm_R_assault <- lm_robust(spv_Rassault ~ condition_dummy * PID_bin,
                          data = data_clean, se_type = "HC2")
# D: Individual SPV (Dems only)
lm_D_assault <- lm_robust(spv_Dassault ~ condition_dummy * PID_bin,
                          data = data_clean, se_type = "HC2")

Individual SPV

term	estimate	std.error	statistic	p.value	conf.low	conf.high	df	outcome
(Intercept)	9.179	2.423	3.788	0.000	4.422	13.936	776	SPV_indiv_assault
condition_dummy	10.027	4.814	2.083	0.038	0.577	19.476	776	SPV_indiv_assault
PID_dummy	-12.890	3.770	-3.419	0.001	-20.291	-5.489	776	SPV_indiv_assault
PID_bin	1.572	1.618	0.972	0.331	-1.603	4.748	776	SPV_indiv_assault
condition_dummy:PID_dummy	-8.759	6.489	-1.350	0.177	-21.498	3.980	776	SPV_indiv_assault
condition_dummy:PID_bin	-5.673	2.875	-1.973	0.049	-11.316	-0.029	776	SPV_indiv_assault
PID_dummy:PID_bin	10.652	3.038	3.506	0.000	4.687	16.616	776	SPV_indiv_assault
condition_dummy:PID_dummy:PID_bin	4.767	5.008	0.952	0.341	-5.063	14.597	776	SPV_indiv_assault

Meta-Perceptions Republican SPV

term	estimate	std.error	statistic	p.value	conf.low	conf.high	df	outcome
(Intercept)	32.391	4.947	6.548	0.000	22.681	42.102	776	SPV_meta_assault
condition_dummy	18.053	7.672	2.353	0.019	2.991	33.114	776	SPV_meta_assault
PID_dummy	-32.920	5.765	-5.711	0.000	-44.237	-21.604	776	SPV_meta_assault
PID_bin	0.344	3.010	0.114	0.909	-5.565	6.252	776	SPV_meta_assault
condition_dummy:PID_dummy	7.925	10.202	0.777	0.438	-12.101	27.951	776	SPV_meta_assault
condition_dummy:PID_bin	-7.579	4.616	-1.642	0.101	-16.641	1.483	776	SPV_meta_assault
PID_dummy:PID_bin	14.458	3.898	3.709	0.000	6.805	22.111	776	SPV_meta_assault
condition_dummy:PID_dummy:PID_bin	-7.131	6.747	-1.057	0.291	-20.376	6.114	776	SPV_meta_assault

Individual SPV (Republicans Only)

term	estimate	std.error	statistic	p.value	conf.low	conf.high	df	outcome
(Intercept)	-3.711	2.888	-1.285	0.200	-9.399	1.976	254	spv_Rassault
condition_dummy	1.268	4.352	0.291	0.771	-7.303	9.838	254	spv_Rassault
PID_bin	12.224	2.572	4.753	0.000	7.159	17.289	254	spv_Rassault
condition_dummy:PID_bin	-0.906	4.100	-0.221	0.825	-8.981	7.169	254	spv_Rassault

Individual SPV (Democrats Only)

term	estimate	std.error	statistic	p.value	conf.low	conf.high	df	outcome
(Intercept)	9.179	2.423	3.788	0.000	4.418	13.940	522	spv_Dassault
condition_dummy	10.027	4.814	2.083	0.038	0.570	19.484	522	spv_Dassault
PID_bin	1.572	1.618	0.972	0.331	-1.605	4.750	522	spv_Dassault
condition_dummy:PID_bin	-5.673	2.875	-1.973	0.049	-11.320	-0.025	522	spv_Dassault

Party ID (Dummy)

# form emmeans and contrast matrices for Individual SPV (Aggregated)
emm_assault_indiv <- emmeans(lm_assault_indiv, ~ condition_dummy * PID_dummy,
                             vcov. = vcov(lm_assault_indiv))
emm_assault_indiv_df <- as.data.frame(confint(emm_assault_indiv)) %>%
  mutate(
    condition_dummy = factor(condition_dummy, levels = c(0, 1),
                             labels = c("Control", "Treatment")),
    PID_dummy = factor(PID_dummy, levels = c(0, 1),
                       labels = c("Democrat", "Republican")),
    group = interaction(condition_dummy, PID_dummy))
contr_assault_indiv <- contrast(emm_assault_indiv,
                                method = list(
                                  "Control D vs. Control R"     = c(-1,  0,  1,  0),
                                  "Control D vs. Treatment D"   = c(-1,  1,  0,  0),
                                  "Control R vs. Treatment R"   = c( 0,  0, -1,  1),
                                  "Treatment D vs. Treatment R" = c( 0, -1,  0,  1)))
contr_assault_indiv_df <- as.data.frame(confint(contr_assault_indiv))
contr_table_assault_indiv <- summary(contr_assault_indiv) |>
  as.data.frame() |>
  dplyr::select(contrast, estimate, SE, df, t.ratio, p.value)

contrast	estimate	SE	df	t.ratio	p.value
Control D vs. Control R	3.161920	2.421940	776	1.3055317	0.1920988
Control D vs. Treatment D	1.477989	1.862313	776	0.7936311	0.4276528
Control R vs. Treatment R	-0.097438	3.235799	776	-0.0301125	0.9759851
Treatment D vs. Treatment R	1.586492	2.841269	776	0.5583746	0.5767497

# form emmeans and contrast matrices for Meta-Perceptions of Republican SPV
emm_assault_meta_R <- emmeans(lm_assault_meta_R, ~ condition_dummy * PID_dummy,
                              vcov. = vcov(lm_assault_meta_R))
emm_assault_meta_R_df <- as.data.frame(confint(emm_assault_meta_R)) %>%
  mutate(
    condition_dummy = factor(condition_dummy, levels = c(0, 1),
                             labels = c("Control", "Treatment")),
    PID_dummy = factor(PID_dummy, levels = c(0, 1),
                       labels = c("Democrat", "Republican")),
    group = interaction(condition_dummy, PID_dummy))
contr_assault_meta_R <- contrast(emm_assault_meta_R,
                                 method = list(
                                   "Control D vs. Control R"     = c(-1,  0,  1,  0),
                                   "Control D vs. Treatment D"   = c(-1,  1,  0,  0),
                                   "Control R vs. Treatment R"   = c( 0,  0, -1,  1),
                                   "Treatment D vs. Treatment R" = c( 0, -1,  0,  1)))
contr_assault_meta_R_df <- as.data.frame(confint(contr_assault_meta_R))
contr_table_assault_meta_R <- summary(contr_assault_meta_R) |>
  as.data.frame() |>
  dplyr::select(contrast, estimate, SE, df, t.ratio, p.value)

contrast	estimate	SE	df	t.ratio	p.value
Control D vs. Control R	-11.131716	2.804346	776	-3.969451	0.0000787
Control D vs. Treatment D	6.631068	2.886408	776	2.297343	0.0218646
Control R vs. Treatment R	3.809534	3.341862	776	1.139943	0.2546615
Treatment D vs. Treatment R	-13.953250	3.411016	776	-4.090643	0.0000475

Party ID (7-pt)

# form emmeans and contrast matrices for Individual SPV (Aggregated)
assault_indiv_partyaffil <- lm_robust(SPV_indiv_assault ~ condition_dummy * party_numeric, data = data_clean, se_type = "HC2")
emm_assault_indiv_partyaffil<- emmeans(assault_indiv_partyaffil, ~ condition_dummy * party_numeric,
                                       at = list(party_numeric = c(1, 2, 3, 5, 6, 7)),
                                       vcov. = vcov(assault_indiv_partyaffil))
contr_assault_indiv_partyaffil<- contrast(emm_assault_indiv_partyaffil,
                                          method = list("Treatment - Control" = c(-1, 1)),
                                          by = "party_numeric")
contr_assault_indiv_partyaffil_df <- confint(contr_assault_indiv_partyaffil) |>
  as.data.frame() |>
  mutate(
    party_label = factor(party_numeric, levels = c(1, 2, 3, 5, 6, 7),
                         labels = c("Strong D","Weak D","Lean D","Lean R","Weak R","Strong R")),
    model = "Individual Ratings")

contrast	party_numeric	estimate	SE	df	lower.CL	upper.CL	party_label	model
Treatment - Control	1	0.8867638	2.117820	784	-3.270506	5.044033	Strong D	Individual Ratings
Treatment - Control	2	0.7101517	1.697449	784	-2.621932	4.042235	Weak D	Individual Ratings
Treatment - Control	3	0.5335396	1.602081	784	-2.611337	3.678416	Lean D	Individual Ratings
Treatment - Control	5	0.1803155	2.409315	784	-4.549156	4.909787	Lean R	Individual Ratings
Treatment - Control	6	0.0037034	3.059020	784	-6.001136	6.008542	Weak R	Individual Ratings
Treatment - Control	7	-0.1729087	3.768231	784	-7.569925	7.224108	Strong R	Individual Ratings

# form emmeans and contrast matrices for Meta-Perceptions of Republican SPV
assault_meta_R_partyaffil <- lm_robust(SPV_meta_assault ~ condition_dummy * party_numeric, data = data_clean, se_type = "HC2")
emm_assault_meta_R_partyaffil<- emmeans(assault_meta_R_partyaffil, ~ condition_dummy * party_numeric,
                                        at = list(party_numeric = c(1, 2, 3, 5, 6, 7)),
                                        vcov. = vcov(assault_meta_R_partyaffil))
contr_assault_meta_R_partyaffil<- contrast(emm_assault_meta_R_partyaffil,
                                           method = list("Treatment - Control" = c(-1, 1)),
                                           by = "party_numeric")
contr_assault_meta_R_partyaffil_df <- confint(contr_assault_meta_R_partyaffil) |>
  as.data.frame() |>
  mutate(
    party_label = factor(party_numeric, levels = c(1, 2, 3, 5, 6, 7),
                         labels = c("Strong D","Weak D","Lean D","Lean R","Weak R","Strong R")),
    model = "Meta-Perceptions of Republicans' Ratings")

contrast	party_numeric	estimate	SE	df	lower.CL	upper.CL	party_label	model
Treatment - Control	1	6.562129	3.253665	784	0.1752020	12.94906	Strong D	Meta-Perceptions of Republicans’ Ratings
Treatment - Control	2	6.167428	2.634814	784	0.9953029	11.33955	Weak D	Meta-Perceptions of Republicans’ Ratings
Treatment - Control	3	5.772727	2.257687	784	1.3408998	10.20455	Lean D	Meta-Perceptions of Republicans’ Ratings
Treatment - Control	5	4.983325	2.608441	784	-0.1370299	10.10368	Lean R	Meta-Perceptions of Republicans’ Ratings
Treatment - Control	6	4.588624	3.218054	784	-1.7283977	10.90565	Weak R	Meta-Perceptions of Republicans’ Ratings
Treatment - Control	7	4.193923	3.963178	784	-3.5857725	11.97362	Strong R	Meta-Perceptions of Republicans’ Ratings

Party ID (7-pt) + Partisan Self-Concept

lm_assault_indiv_R_7 <- lm_robust(SPV_indiv_assault ~ condition_dummy * party_numeric * PID_bin,
                                  data = data_clean, se_type = "HC2")
emm_assault_indiv_R_7<- emmeans(lm_assault_indiv_R_7, ~ condition_dummy | party_numeric,
                                at = list(party_numeric = c(1, 2, 3, 5, 6, 7)),
                                vcov. = vcov(lm_assault_indiv_R_7))
contr_assault_indiv_R_7<- contrast(emm_assault_indiv_R_7,
                                   method = list("Treatment - Control" = c(-1, 1)),
                                   by = "party_numeric")
contr_assault_indiv_R_7_df <- as.data.frame(confint(contr_assault_indiv_R_7))
contr_table_assault_indiv_R_7_df <- summary(contr_assault_indiv_R_7) |>
  as.data.frame() |>
  dplyr::select(contrast, party_numeric, estimate, SE, df, t.ratio, p.value) |>
  mutate(party_label = factor(party_numeric, levels = c(1, 2, 3, 5, 6, 7),
                              labels = c("Strong D","Weak D","Lean D","Lean R","Weak R","Strong R")))

contrast	party_numeric	estimate	SE	df	t.ratio	p.value	party_label
Treatment - Control	1	1.7998829	2.238053	776	0.8042181	0.4215173	Strong D
Treatment - Control	2	1.5032622	1.829128	776	0.8218464	0.4114168	Weak D
Treatment - Control	3	1.2066415	1.683575	776	0.7167137	0.4737664	Lean D
Treatment - Control	5	0.6134002	2.295110	776	0.2672640	0.7893369	Lean R
Treatment - Control	6	0.3167795	2.865644	776	0.1105439	0.9120066	Weak R
Treatment - Control	7	0.0201588	3.508303	776	0.0057460	0.9954168	Strong R

lm_assault_meta_R_7 <- lm_robust(SPV_meta_assault ~ condition_dummy * party_numeric * PID_bin,
                                 data = data_clean, se_type = "HC2")
emm_assault_meta_R_7<- emmeans(lm_assault_meta_R_7, ~ condition_dummy | party_numeric,
                               at = list(party_numeric = c(1, 2, 3, 5, 6, 7)),
                               vcov. = vcov(lm_assault_meta_R_7))
contr_assault_meta_R_7<- contrast(emm_assault_meta_R_7,
                                  method = list("Treatment - Control" = c(-1, 1)),
                                  by = "party_numeric")
contr_assault_meta_R_7_df <- as.data.frame(confint(contr_assault_meta_R_7))
contr_table_assault_meta_R_7_df <- summary(contr_assault_meta_R_7) |>
  as.data.frame() |>
  dplyr::select(contrast, party_numeric, estimate, SE, df, t.ratio, p.value) |>
  mutate(party_label = factor(party_numeric, levels = c(1, 2, 3, 5, 6, 7),
                              labels = c("Strong D","Weak D","Lean D","Lean R","Weak R","Strong R")))

contrast	party_numeric	estimate	SE	df	t.ratio	p.value	party_label
Treatment - Control	1	6.604987	3.443888	776	1.917887	0.0554921	Strong D
Treatment - Control	2	6.283123	2.774568	776	2.264541	0.0238159	Weak D
Treatment - Control	3	5.961260	2.317901	776	2.571836	0.0103011	Lean D
Treatment - Control	5	5.317533	2.496252	776	2.130207	0.0334685	Lean R
Treatment - Control	6	4.995669	3.068459	776	1.628071	0.1039157	Weak R
Treatment - Control	7	4.673806	3.799498	776	1.230111	0.2190282	Strong R

Factor models for panel plots

assault_indiv_factor <- lm_robust(SPV_indiv_assault ~ condition_dummy * factor(party_numeric),
                                  data = data_clean, se_type = "HC2")
emm_assault_indiv_factor <- emmeans(assault_indiv_factor, ~ condition_dummy | party_numeric,
                                    vcov. = vcov(assault_indiv_factor))
contr_assault_indiv_factor_df <- confint(contrast(emm_assault_indiv_factor,
                                                  method = list("Treatment - Control" = c(-1, 1)),
                                                  by = "party_numeric")) |>
  as.data.frame() |>
  mutate(party_numeric = as.numeric(as.character(party_numeric)))

assault_meta_R_factor <- lm_robust(SPV_meta_assault ~ condition_dummy * factor(party_numeric),
                                   data = data_clean, se_type = "HC2")
emm_assault_meta_R_factor <- emmeans(assault_meta_R_factor, ~ condition_dummy | party_numeric,
                                     vcov. = vcov(assault_meta_R_factor))
contr_assault_meta_R_factor_df <- confint(contrast(emm_assault_meta_R_factor,
                                                   method = list("Treatment - Control" = c(-1, 1)),
                                                   by = "party_numeric")) |>
  as.data.frame() |>
  mutate(party_numeric = as.numeric(as.character(party_numeric)))

p_assault_indiv <- make_SPV_panels(
  "SPV_indiv_assault",
  contr_assault_indiv_partyaffil_df,
  contr_assault_indiv_factor_df,
  "Individual SPV: Assault\n(raw mean)",
  "Figure: Physically Assaulting an Out-Partisan — Individual Ratings",
  color_A = c("Control R" = "forestgreen", "Treatment R" = "yellowgreen",
              "Control D" = "limegreen",    "Treatment D" = "greenyellow"))

p_assault_meta_R <- make_SPV_panels(
  "SPV_meta_assault",
  contr_assault_meta_R_partyaffil_df,
  contr_assault_meta_R_factor_df,
  "Meta-Perceptions of Republican SPV: Assault\n(raw mean)",
  "Figure: Physically Assaulting an Out-Partisan — Meta-Perceptions",
  color_A = c("Control R" = "forestgreen", "Treatment R" = "yellowgreen",
              "Control D" = "limegreen",    "Treatment D" = "greenyellow"))

Scenario: To achieve political goals

Models

# 1.1: form models for:
# A: Individual SPV (Aggregated)
lm_justify_indiv <- lm_robust(SPV_indiv_justify ~ condition_dummy * PID_dummy * PID_bin,
                              data = data_clean, se_type = "HC2")
# B: Meta-Perceptions of Republican SPV 
lm_justify_meta_R <- lm_robust(SPV_meta_justify ~ condition_dummy * PID_dummy * PID_bin,
                               data = data_clean, se_type = "HC2")
# C: Individual SPV (Reps only)
lm_R_justify <- lm_robust(spv_Rjustify ~ condition_dummy * PID_bin,
                          data = data_clean, se_type = "HC2")
# D: Individual SPV (Dems only)
lm_D_justify <- lm_robust(spv_Djustify ~ condition_dummy * PID_bin,
                          data = data_clean, se_type = "HC2")

Individual SPV

term	estimate	std.error	statistic	p.value	conf.low	conf.high	df	outcome
(Intercept)	14.413	3.166	4.553	0.000	8.198	20.627	777	SPV_indiv_justify
condition_dummy	8.146	5.645	1.443	0.149	-2.936	19.228	777	SPV_indiv_justify
PID_dummy	-10.883	6.054	-1.798	0.073	-22.768	1.001	777	SPV_indiv_justify
PID_bin	0.348	1.954	0.178	0.859	-3.488	4.184	777	SPV_indiv_justify
condition_dummy:PID_dummy	-13.252	8.346	-1.588	0.113	-29.634	3.131	777	SPV_indiv_justify
condition_dummy:PID_bin	-3.449	3.395	-1.016	0.310	-10.113	3.215	777	SPV_indiv_justify
PID_dummy:PID_bin	7.190	3.988	1.803	0.072	-0.638	15.018	777	SPV_indiv_justify
condition_dummy:PID_dummy:PID_bin	5.163	5.825	0.886	0.376	-6.272	16.598	777	SPV_indiv_justify

Meta-Perceptions Republican SPV

term	estimate	std.error	statistic	p.value	conf.low	conf.high	df	outcome
(Intercept)	41.347	5.020	8.237	0.000	31.493	51.201	777	SPV_meta_justify
condition_dummy	2.817	7.838	0.359	0.719	-12.568	18.202	777	SPV_meta_justify
PID_dummy	-29.849	6.339	-4.709	0.000	-42.293	-17.405	777	SPV_meta_justify
PID_bin	-1.137	3.082	-0.369	0.712	-7.188	4.913	777	SPV_meta_justify
condition_dummy:PID_dummy	11.064	10.896	1.015	0.310	-10.324	32.452	777	SPV_meta_justify
condition_dummy:PID_bin	0.982	4.728	0.208	0.835	-8.300	10.264	777	SPV_meta_justify
PID_dummy:PID_bin	7.981	4.150	1.923	0.055	-0.166	16.127	777	SPV_meta_justify
condition_dummy:PID_dummy:PID_bin	-9.722	6.998	-1.389	0.165	-23.459	4.015	777	SPV_meta_justify

Individual SPV (Republicans Only)

term	estimate	std.error	statistic	p.value	conf.low	conf.high	df	outcome
(Intercept)	3.529	5.161	0.684	0.495	-6.634	13.692	255	spv_Rjustify
condition_dummy	-5.106	6.146	-0.831	0.407	-17.210	6.998	255	spv_Rjustify
PID_bin	7.538	3.476	2.169	0.031	0.693	14.383	255	spv_Rjustify
condition_dummy:PID_bin	1.714	4.734	0.362	0.718	-7.608	11.036	255	spv_Rjustify

Individual SPV (Democrats Only)

term	estimate	std.error	statistic	p.value	conf.low	conf.high	df	outcome
(Intercept)	14.413	3.166	4.553	0.000	8.193	20.632	522	spv_Djustify
condition_dummy	8.146	5.645	1.443	0.150	-2.945	19.236	522	spv_Djustify
PID_bin	0.348	1.954	0.178	0.859	-3.491	4.187	522	spv_Djustify
condition_dummy:PID_bin	-3.449	3.395	-1.016	0.310	-10.118	3.220	522	spv_Djustify

Party ID (Dummy)

# form emmeans and contrast matrices for Individual SPV (Aggregated)
emm_justify_indiv <- emmeans(lm_justify_indiv, ~ condition_dummy * PID_dummy,
                             vcov. = vcov(lm_justify_indiv))
emm_justify_indiv_df <- as.data.frame(confint(emm_justify_indiv)) %>%
  mutate(
    condition_dummy = factor(condition_dummy, levels = c(0, 1),
                             labels = c("Control", "Treatment")),
    PID_dummy = factor(PID_dummy, levels = c(0, 1),
                       labels = c("Democrat", "Republican")),
    group = interaction(condition_dummy, PID_dummy))
contr_justify_indiv <- contrast(emm_justify_indiv,
                                method = list(
                                  "Control D vs. Control R"     = c(-1,  0,  1,  0),
                                  "Control D vs. Treatment D"   = c(-1,  1,  0,  0),
                                  "Control R vs. Treatment R"   = c( 0,  0, -1,  1),
                                  "Treatment D vs. Treatment R" = c( 0, -1,  0,  1)))
contr_justify_indiv_df <- as.data.frame(confint(contr_justify_indiv))
contr_table_justify_indiv <- summary(contr_justify_indiv) |>
  as.data.frame() |>
  dplyr::select(contrast, estimate, SE, df, t.ratio, p.value)

contrast	estimate	SE	df	t.ratio	p.value
Control D vs. Control R	-0.0481568	2.583281	777	-0.0186417	0.9851317
Control D vs. Treatment D	2.9481517	2.129194	777	1.3846327	0.1665623
Control R vs. Treatment R	-2.5234938	3.168450	777	-0.7964442	0.4260171
Treatment D vs. Treatment R	-5.5198023	2.810552	777	-1.9639568	0.0498912

# form emmeans and contrast matrices for Meta-Perceptions of Republican SPV
emm_justify_meta_R <- emmeans(lm_justify_meta_R, ~ condition_dummy * PID_dummy,
                              vcov. = vcov(lm_justify_meta_R))
emm_justify_meta_R_df <- as.data.frame(confint(emm_justify_meta_R)) %>%
  mutate(
    condition_dummy = factor(condition_dummy, levels = c(0, 1),
                             labels = c("Control", "Treatment")),
    PID_dummy = factor(PID_dummy, levels = c(0, 1),
                       labels = c("Democrat", "Republican")),
    group = interaction(condition_dummy, PID_dummy))
contr_justify_meta_R <- contrast(emm_justify_meta_R,
                                 method = list(
                                   "Control D vs. Control R"     = c(-1,  0,  1,  0),
                                   "Control D vs. Treatment D"   = c(-1,  1,  0,  0),
                                   "Control R vs. Treatment R"   = c( 0,  0, -1,  1),
                                   "Treatment D vs. Treatment R" = c( 0, -1,  0,  1)))
contr_justify_meta_R_df <- as.data.frame(confint(contr_justify_meta_R))
contr_table_justify_meta_R <- summary(contr_justify_meta_R) |>
  as.data.frame() |>
  dplyr::select(contrast, estimate, SE, df, t.ratio, p.value)

contrast	estimate	SE	df	t.ratio	p.value
Control D vs. Control R	-17.822025	2.854089	777	-6.2443824	0.0000000
Control D vs. Treatment D	4.297555	2.926006	777	1.4687444	0.1423069
Control R vs. Treatment R	0.710639	3.206779	777	0.2216052	0.8246794
Treatment D vs. Treatment R	-21.408941	3.270951	777	-6.5451737	0.0000000

Party ID (7-pt)

# form emmeans and contrast matrices for Individual SPV (Aggregated)
justify_indiv_partyaffil <- lm_robust(SPV_indiv_justify ~ condition_dummy * party_numeric, data = data_clean, se_type = "HC2")
emm_justify_indiv_partyaffil<- emmeans(justify_indiv_partyaffil, ~ condition_dummy * party_numeric,
                                       at = list(party_numeric = c(1, 2, 3, 5, 6, 7)),
                                       vcov. = vcov(justify_indiv_partyaffil))
contr_justify_indiv_partyaffil<- contrast(emm_justify_indiv_partyaffil,
                                          method = list("Treatment - Control" = c(-1, 1)),
                                          by = "party_numeric")
contr_justify_indiv_partyaffil_df <- confint(contr_justify_indiv_partyaffil) |>
  as.data.frame() |>
  mutate(
    party_label = factor(party_numeric, levels = c(1, 2, 3, 5, 6, 7),
                         labels = c("Strong D","Weak D","Lean D","Lean R","Weak R","Strong R")),
    model = "Individual Ratings")

contrast	party_numeric	estimate	SE	df	lower.CL	upper.CL	party_label	model
Treatment - Control	1	2.5208248	2.394745	785	-2.180038	7.221687	Strong D	Individual Ratings
Treatment - Control	2	1.8361252	1.938295	785	-1.968729	5.640979	Weak D	Individual Ratings
Treatment - Control	3	1.1514257	1.745909	785	-2.275776	4.578628	Lean D	Individual Ratings
Treatment - Control	5	-0.2179735	2.331890	785	-4.795451	4.359504	Lean R	Individual Ratings
Treatment - Control	6	-0.9026731	2.921645	785	-6.637835	4.832489	Weak R	Individual Ratings
Treatment - Control	7	-1.5873726	3.592122	785	-8.638675	5.463930	Strong R	Individual Ratings

# form emmeans and contrast matrices for Meta-Perceptions of Republican SPV
justify_meta_R_partyaffil <- lm_robust(SPV_meta_justify ~ condition_dummy * party_numeric, data = data_clean, se_type = "HC2")
emm_justify_meta_R_partyaffil<- emmeans(justify_meta_R_partyaffil, ~ condition_dummy * party_numeric,
                                        at = list(party_numeric = c(1, 2, 3, 5, 6, 7)),
                                        vcov. = vcov(justify_meta_R_partyaffil))
contr_justify_meta_R_partyaffil<- contrast(emm_justify_meta_R_partyaffil,
                                           method = list("Treatment - Control" = c(-1, 1)),
                                           by = "party_numeric")
contr_justify_meta_R_partyaffil_df <- confint(contr_justify_meta_R_partyaffil) |>
  as.data.frame() |>
  mutate(
    party_label = factor(party_numeric, levels = c(1, 2, 3, 5, 6, 7),
                         labels = c("Strong D","Weak D","Lean D","Lean R","Weak R","Strong R")),
    model = "Meta-Perceptions of Republicans' Ratings")

contrast	party_numeric	estimate	SE	df	lower.CL	upper.CL	party_label	model
Treatment - Control	1	5.133102	3.298748	785	-1.3423106	11.608514	Strong D	Meta-Perceptions of Republicans’ Ratings
Treatment - Control	2	4.458389	2.673551	785	-0.7897656	9.706544	Weak D	Meta-Perceptions of Republicans’ Ratings
Treatment - Control	3	3.783677	2.258464	785	-0.6496661	8.217021	Lean D	Meta-Perceptions of Republicans’ Ratings
Treatment - Control	5	2.434253	2.462964	785	-2.4005216	7.269027	Lean R	Meta-Perceptions of Republicans’ Ratings
Treatment - Control	6	1.759541	3.013127	785	-4.1551996	7.674281	Weak R	Meta-Perceptions of Republicans’ Ratings
Treatment - Control	7	1.084828	3.711918	785	-6.2016314	8.371288	Strong R	Meta-Perceptions of Republicans’ Ratings

Party ID (7-pt) + Partisan Self-Concept

lm_justify_indiv_R_7 <- lm_robust(SPV_indiv_justify ~ condition_dummy * party_numeric * PID_bin,
                                  data = data_clean, se_type = "HC2")
emm_justify_indiv_R_7<- emmeans(lm_justify_indiv_R_7, ~ condition_dummy | party_numeric,
                                at = list(party_numeric = c(1, 2, 3, 5, 6, 7)),
                                vcov. = vcov(lm_justify_indiv_R_7))
contr_justify_indiv_R_7<- contrast(emm_justify_indiv_R_7,
                                   method = list("Treatment - Control" = c(-1, 1)),
                                   by = "party_numeric")
contr_justify_indiv_R_7_df <- as.data.frame(confint(contr_justify_indiv_R_7))
contr_table_justify_indiv_R_7_df <- summary(contr_justify_indiv_R_7) |>
  as.data.frame() |>
  dplyr::select(contrast, party_numeric, estimate, SE, df, t.ratio, p.value) |>
  mutate(party_label = factor(party_numeric, levels = c(1, 2, 3, 5, 6, 7),
                              labels = c("Strong D","Weak D","Lean D","Lean R","Weak R","Strong R")))

contrast	party_numeric	estimate	SE	df	t.ratio	p.value	party_label
Treatment - Control	1	3.2048173	2.586395	777	1.2391060	0.2156804	Strong D
Treatment - Control	2	2.3796634	2.093036	777	1.1369435	0.2559124	Weak D
Treatment - Control	3	1.5545095	1.834173	777	0.8475260	0.3969630	Lean D
Treatment - Control	5	-0.0957982	2.281901	777	-0.0419818	0.9665240	Lean R
Treatment - Control	6	-0.9209521	2.840169	777	-0.3242596	0.7458288	Weak R
Treatment - Control	7	-1.7461060	3.495437	777	-0.4995387	0.6175414	Strong R

lm_justify_meta_R_7 <- lm_robust(SPV_meta_justify ~ condition_dummy * party_numeric * PID_bin,
                                 data = data_clean, se_type = "HC2")
emm_justify_meta_R_7<- emmeans(lm_justify_meta_R_7, ~ condition_dummy | party_numeric,
                               at = list(party_numeric = c(1, 2, 3, 5, 6, 7)),
                               vcov. = vcov(lm_justify_meta_R_7))
contr_justify_meta_R_7<- contrast(emm_justify_meta_R_7,
                                  method = list("Treatment - Control" = c(-1, 1)),
                                  by = "party_numeric")
contr_justify_meta_R_7_df <- as.data.frame(confint(contr_justify_meta_R_7))
contr_table_justify_meta_R_7_df <- summary(contr_justify_meta_R_7) |>
  as.data.frame() |>
  dplyr::select(contrast, party_numeric, estimate, SE, df, t.ratio, p.value) |>
  mutate(party_label = factor(party_numeric, levels = c(1, 2, 3, 5, 6, 7),
                              labels = c("Strong D","Weak D","Lean D","Lean R","Weak R","Strong R")))

contrast	party_numeric	estimate	SE	df	t.ratio	p.value	party_label
Treatment - Control	1	3.970401	3.486615	777	1.1387552	0.2551562	Strong D
Treatment - Control	2	3.687270	2.802397	777	1.3157557	0.1886441	Weak D
Treatment - Control	3	3.404140	2.317908	777	1.4686262	0.1423389	Lean D
Treatment - Control	5	2.837879	2.424495	777	1.1705031	0.2421574	Lean R
Treatment - Control	6	2.554748	2.977313	777	0.8580719	0.3911174	Weak R
Treatment - Control	7	2.271618	3.697692	777	0.6143340	0.5391745	Strong R

Factor models for panel plots

justify_indiv_factor <- lm_robust(SPV_indiv_justify ~ condition_dummy * factor(party_numeric),
                                  data = data_clean, se_type = "HC2")
emm_justify_indiv_factor <- emmeans(justify_indiv_factor, ~ condition_dummy | party_numeric,
                                    vcov. = vcov(justify_indiv_factor))
contr_justify_indiv_factor_df <- confint(contrast(emm_justify_indiv_factor,
                                                  method = list("Treatment - Control" = c(-1, 1)),
                                                  by = "party_numeric")) |>
  as.data.frame() |>
  mutate(party_numeric = as.numeric(as.character(party_numeric)))

justify_meta_R_factor <- lm_robust(SPV_meta_justify ~ condition_dummy * factor(party_numeric),
                                   data = data_clean, se_type = "HC2")
emm_justify_meta_R_factor <- emmeans(justify_meta_R_factor, ~ condition_dummy | party_numeric,
                                     vcov. = vcov(justify_meta_R_factor))
contr_justify_meta_R_factor_df <- confint(contrast(emm_justify_meta_R_factor,
                                                   method = list("Treatment - Control" = c(-1, 1)),
                                                   by = "party_numeric")) |>
  as.data.frame() |>
  mutate(party_numeric = as.numeric(as.character(party_numeric)))

p_justify_indiv <- make_SPV_panels(
  "SPV_indiv_justify",
  contr_justify_indiv_partyaffil_df,
  contr_justify_indiv_factor_df,
  "Individual SPV: Justify\n(raw mean)",
  "Figure: Achieving In-Party Political Goals — Individual Ratings",
  color_A = c("Control R" = "midnightblue", "Treatment R" = "deepskyblue",
              "Control D" = "dodgerblue",    "Treatment D" = "lightblue"))

p_justify_meta_R <- make_SPV_panels(
  "SPV_meta_justify",
  contr_justify_meta_R_partyaffil_df,
  contr_justify_meta_R_factor_df,
  "Meta-Perceptions of Republican SPV: Justify\n(raw mean)",
  "Figure: Achieving In-Party Political Goals — Meta-Perceptions",
  color_A = c("Control R" = "midnightblue", "Treatment R" = "deepskyblue",
              "Control D" = "dodgerblue",    "Treatment D" = "lightblue"))

Scenario: If the other party wins the next election

Models

# 1.1: form models for:
# A: Individual SPV (Aggregated)
lm_predict_indiv <- lm_robust(SPV_indiv_predict ~ condition_dummy * PID_dummy * PID_bin,
                              data = data_clean, se_type = "HC2")
# B: Meta-Perceptions of Republican SPV 
lm_predict_meta_R <- lm_robust(SPV_meta_predict ~ condition_dummy * PID_dummy * PID_bin,
                               data = data_clean, se_type = "HC2")
# C: Individual SPV (Reps only)
lm_R_predict <- lm_robust(spv_Rpredict ~ condition_dummy * PID_bin,
                          data = data_clean, se_type = "HC2")
# D: Individual SPV (Dems only)
lm_D_predict <- lm_robust(spv_Dpredict ~ condition_dummy * PID_bin,
                          data = data_clean, se_type = "HC2")

Individual SPV

term	estimate	std.error	statistic	p.value	conf.low	conf.high	df	outcome
(Intercept)	8.684	2.738	3.171	0.002	3.309	14.060	774	SPV_indiv_predict
condition_dummy	17.457	5.768	3.027	0.003	6.135	28.780	774	SPV_indiv_predict
PID_dummy	-10.176	4.263	-2.387	0.017	-18.544	-1.808	774	SPV_indiv_predict
PID_bin	2.585	1.848	1.399	0.162	-1.042	6.212	774	SPV_indiv_predict
condition_dummy:PID_dummy	-19.921	7.012	-2.841	0.005	-33.686	-6.155	774	SPV_indiv_predict
condition_dummy:PID_bin	-8.557	3.462	-2.471	0.014	-15.353	-1.760	774	SPV_indiv_predict
PID_dummy:PID_bin	5.563	3.183	1.748	0.081	-0.684	11.811	774	SPV_indiv_predict
condition_dummy:PID_dummy:PID_bin	9.284	5.019	1.850	0.065	-0.569	19.137	774	SPV_indiv_predict

Meta-Perceptions Republican SPV

term	estimate	std.error	statistic	p.value	conf.low	conf.high	df	outcome
(Intercept)	39.627	5.498	7.207	0.000	28.834	50.420	775	SPV_meta_predict
condition_dummy	10.335	7.970	1.297	0.195	-5.311	25.981	775	SPV_meta_predict
PID_dummy	-33.162	6.271	-5.288	0.000	-45.472	-20.851	775	SPV_meta_predict
PID_bin	2.282	3.347	0.682	0.496	-4.289	8.852	775	SPV_meta_predict
condition_dummy:PID_dummy	4.300	10.272	0.419	0.676	-15.864	24.464	775	SPV_meta_predict
condition_dummy:PID_bin	-3.871	4.864	-0.796	0.426	-13.418	5.677	775	SPV_meta_predict
PID_dummy:PID_bin	5.689	4.083	1.393	0.164	-2.327	13.705	775	SPV_meta_predict
condition_dummy:PID_dummy:PID_bin	-4.508	6.656	-0.677	0.498	-17.574	8.559	775	SPV_meta_predict

Individual SPV (Republicans Only)

term	estimate	std.error	statistic	p.value	conf.low	conf.high	df	outcome
(Intercept)	-1.492	3.267	-0.457	0.648	-7.926	4.942	255	spv_Rpredict
condition_dummy	-2.463	3.988	-0.618	0.537	-10.317	5.390	255	spv_Rpredict
PID_bin	8.148	2.591	3.145	0.002	3.045	13.251	255	spv_Rpredict
condition_dummy:PID_bin	0.728	3.634	0.200	0.841	-6.429	7.884	255	spv_Rpredict

Individual SPV (Democrats Only)

term	estimate	std.error	statistic	p.value	conf.low	conf.high	df	outcome
(Intercept)	8.684	2.738	3.171	0.002	3.305	14.064	519	spv_Dpredict
condition_dummy	17.457	5.768	3.027	0.003	6.126	28.789	519	spv_Dpredict
PID_bin	2.585	1.848	1.399	0.162	-1.045	6.215	519	spv_Dpredict
condition_dummy:PID_bin	-8.557	3.462	-2.471	0.014	-15.358	-1.755	519	spv_Dpredict

Party ID (Dummy)

# form emmeans and contrast matrices for Individual SPV (Aggregated)
emm_predict_indiv <- emmeans(lm_predict_indiv, ~ condition_dummy * PID_dummy,
                             vcov. = vcov(lm_predict_indiv))
emm_predict_indiv_df <- as.data.frame(confint(emm_predict_indiv)) %>%
  mutate(
    condition_dummy = factor(condition_dummy, levels = c(0, 1),
                             labels = c("Control", "Treatment")),
    PID_dummy = factor(PID_dummy, levels = c(0, 1),
                       labels = c("Democrat", "Republican")),
    group = interaction(condition_dummy, PID_dummy))
contr_predict_indiv <- contrast(emm_predict_indiv,
                                method = list(
                                  "Control D vs. Control R"     = c(-1,  0,  1,  0),
                                  "Control D vs. Treatment D"   = c(-1,  1,  0,  0),
                                  "Control R vs. Treatment R"   = c( 0,  0, -1,  1),
                                  "Treatment D vs. Treatment R" = c( 0, -1,  0,  1)))
contr_predict_indiv_df <- as.data.frame(confint(contr_predict_indiv))
contr_table_predict_indiv <- summary(contr_predict_indiv) |>
  as.data.frame() |>
  dplyr::select(contrast, estimate, SE, df, t.ratio, p.value)

contrast	estimate	SE	df	t.ratio	p.value
Control D vs. Control R	-1.791972	2.244967	774	-0.7982174	0.4249892
Control D vs. Treatment D	4.562643	2.103483	774	2.1690898	0.0303796
Control R vs. Treatment R	-1.366805	2.655832	774	-0.5146426	0.6069497
Treatment D vs. Treatment R	-7.721419	2.537363	774	-3.0430884	0.0024208

# form emmeans and contrast matrices for Meta-Perceptions of Republican SPV
emm_predict_meta_R <- emmeans(lm_predict_meta_R, ~ condition_dummy * PID_dummy,
                              vcov. = vcov(lm_predict_meta_R))
emm_predict_meta_R_df <- as.data.frame(confint(emm_predict_meta_R)) %>%
  mutate(
    condition_dummy = factor(condition_dummy, levels = c(0, 1),
                             labels = c("Control", "Treatment")),
    PID_dummy = factor(PID_dummy, levels = c(0, 1),
                       labels = c("Democrat", "Republican")),
    group = interaction(condition_dummy, PID_dummy))
contr_predict_meta_R <- contrast(emm_predict_meta_R,
                                 method = list(
                                   "Control D vs. Control R"     = c(-1,  0,  1,  0),
                                   "Control D vs. Treatment D"   = c(-1,  1,  0,  0),
                                   "Control R vs. Treatment R"   = c( 0,  0, -1,  1),
                                   "Treatment D vs. Treatment R" = c( 0, -1,  0,  1)))
contr_predict_meta_R_df <- as.data.frame(confint(contr_predict_meta_R))
contr_table_predict_meta_R <- summary(contr_predict_meta_R) |>
  as.data.frame() |>
  dplyr::select(contrast, estimate, SE, df, t.ratio, p.value)

contrast	estimate	SE	df	t.ratio	p.value
Control D vs. Control R	-24.588731	2.767836	775	-8.8837392	0.0000000
Control D vs. Treatment D	4.501966	3.006415	775	1.4974534	0.1346825
Control R vs. Treatment R	2.008337	3.010590	775	0.6670909	0.5049127
Treatment D vs. Treatment R	-27.082360	3.231295	775	-8.3812707	0.0000000

Party ID (7-pt)

# form emmeans and contrast matrices for Individual SPV (Aggregated)
predict_indiv_partyaffil <- lm_robust(SPV_indiv_predict ~ condition_dummy * party_numeric, data = data_clean, se_type = "HC2")
emm_predict_indiv_partyaffil<- emmeans(predict_indiv_partyaffil, ~ condition_dummy * party_numeric,
                                       at = list(party_numeric = c(1, 2, 3, 5, 6, 7)),
                                       vcov. = vcov(predict_indiv_partyaffil))
contr_predict_indiv_partyaffil<- contrast(emm_predict_indiv_partyaffil,
                                          method = list("Treatment - Control" = c(-1, 1)),
                                          by = "party_numeric")
contr_predict_indiv_partyaffil_df <- confint(contr_predict_indiv_partyaffil) |>
  as.data.frame() |>
  mutate(
    party_label = factor(party_numeric, levels = c(1, 2, 3, 5, 6, 7),
                         labels = c("Strong D","Weak D","Lean D","Lean R","Weak R","Strong R")),
    model = "Individual Ratings")

contrast	party_numeric	estimate	SE	df	lower.CL	upper.CL	party_label	model
Treatment - Control	1	4.6637554	2.346620	782	0.0573361	9.270175	Strong D	Individual Ratings
Treatment - Control	2	3.6530960	1.905508	782	-0.0874206	7.393613	Weak D	Individual Ratings
Treatment - Control	3	2.6424366	1.660807	782	-0.6177305	5.902604	Lean D	Individual Ratings
Treatment - Control	5	0.6211177	2.005456	782	-3.3155961	4.557832	Lean R	Individual Ratings
Treatment - Control	6	-0.3895417	2.481554	782	-5.2608370	4.481753	Weak R	Individual Ratings
Treatment - Control	7	-1.4002012	3.049155	782	-7.3856985	4.585296	Strong R	Individual Ratings

# form emmeans and contrast matrices for Meta-Perceptions of Republican SPV
predict_meta_R_partyaffil <- lm_robust(SPV_meta_predict ~ condition_dummy * party_numeric, data = data_clean, se_type = "HC2")
emm_predict_meta_R_partyaffil<- emmeans(predict_meta_R_partyaffil, ~ condition_dummy * party_numeric,
                                        at = list(party_numeric = c(1, 2, 3, 5, 6, 7)),
                                        vcov. = vcov(predict_meta_R_partyaffil))
contr_predict_meta_R_partyaffil<- contrast(emm_predict_meta_R_partyaffil,
                                           method = list("Treatment - Control" = c(-1, 1)),
                                           by = "party_numeric")
contr_predict_meta_R_partyaffil_df <- confint(contr_predict_meta_R_partyaffil) |>
  as.data.frame() |>
  mutate(
    party_label = factor(party_numeric, levels = c(1, 2, 3, 5, 6, 7),
                         labels = c("Strong D","Weak D","Lean D","Lean R","Weak R","Strong R")),
    model = "Meta-Perceptions of Republicans' Ratings")

contrast	party_numeric	estimate	SE	df	lower.CL	upper.CL	party_label	model
Treatment - Control	1	4.927844	3.420618	783	-1.7868235	11.642511	Strong D	Meta-Perceptions of Republicans’ Ratings
Treatment - Control	2	4.620863	2.771708	783	-0.8199958	10.061721	Weak D	Meta-Perceptions of Republicans’ Ratings
Treatment - Control	3	4.313882	2.304094	783	-0.2090517	8.836815	Lean D	Meta-Perceptions of Republicans’ Ratings
Treatment - Control	5	3.699919	2.344367	783	-0.9020688	8.301908	Lean R	Meta-Perceptions of Republicans’ Ratings
Treatment - Control	6	3.392938	2.838446	783	-2.1789269	8.964804	Weak R	Meta-Perceptions of Republicans’ Ratings
Treatment - Control	7	3.085957	3.501749	783	-3.7879695	9.959884	Strong R	Meta-Perceptions of Republicans’ Ratings

Party ID (7-pt) + Partisan Self-Concept

lm_predict_indiv_R_7 <- lm_robust(SPV_indiv_predict ~ condition_dummy * party_numeric * PID_bin,
                                  data = data_clean, se_type = "HC2")
emm_predict_indiv_R_7<- emmeans(lm_predict_indiv_R_7, ~ condition_dummy | party_numeric,
                                at = list(party_numeric = c(1, 2, 3, 5, 6, 7)),
                                vcov. = vcov(lm_predict_indiv_R_7))
contr_predict_indiv_R_7<- contrast(emm_predict_indiv_R_7,
                                   method = list("Treatment - Control" = c(-1, 1)),
                                   by = "party_numeric")
contr_predict_indiv_R_7_df <- as.data.frame(confint(contr_predict_indiv_R_7))
contr_table_predict_indiv_R_7_df <- summary(contr_predict_indiv_R_7) |>
  as.data.frame() |>
  dplyr::select(contrast, party_numeric, estimate, SE, df, t.ratio, p.value) |>
  mutate(party_label = factor(party_numeric, levels = c(1, 2, 3, 5, 6, 7),
                              labels = c("Strong D","Weak D","Lean D","Lean R","Weak R","Strong R")))

contrast	party_numeric	estimate	SE	df	t.ratio	p.value	party_label
Treatment - Control	1	6.5198746	2.548158	774	2.5586618	0.0106968	Strong D
Treatment - Control	2	4.9436022	2.063012	774	2.3963035	0.0167978	Weak D
Treatment - Control	3	3.3673299	1.743837	774	1.9309884	0.0538494	Lean D
Treatment - Control	5	0.2147852	1.917551	774	0.1120101	0.9108444	Lean R
Treatment - Control	6	-1.3614872	2.351188	774	-0.5790636	0.5627147	Weak R
Treatment - Control	7	-2.9377596	2.898484	774	-1.0135504	0.3111140	Strong R

lm_predict_meta_R_7 <- lm_robust(SPV_meta_predict ~ condition_dummy * party_numeric * PID_bin,
                                 data = data_clean, se_type = "HC2")
emm_predict_meta_R_7<- emmeans(lm_predict_meta_R_7, ~ condition_dummy | party_numeric,
                               at = list(party_numeric = c(1, 2, 3, 5, 6, 7)),
                               vcov. = vcov(lm_predict_meta_R_7))
contr_predict_meta_R_7<- contrast(emm_predict_meta_R_7,
                                  method = list("Treatment - Control" = c(-1, 1)),
                                  by = "party_numeric")
contr_predict_meta_R_7_df <- as.data.frame(confint(contr_predict_meta_R_7))
contr_table_predict_meta_R_7_df <- summary(contr_predict_meta_R_7) |>
  as.data.frame() |>
  dplyr::select(contrast, party_numeric, estimate, SE, df, t.ratio, p.value) |>
  mutate(party_label = factor(party_numeric, levels = c(1, 2, 3, 5, 6, 7),
                              labels = c("Strong D","Weak D","Lean D","Lean R","Weak R","Strong R")))

contrast	party_numeric	estimate	SE	df	t.ratio	p.value	party_label
Treatment - Control	1	4.468594	3.605082	775	1.2395262	0.2155259	Strong D
Treatment - Control	2	4.297298	2.898957	775	1.4823598	0.1386511	Weak D
Treatment - Control	3	4.126001	2.362126	775	1.7467323	0.0810799	Lean D
Treatment - Control	5	3.783408	2.288245	775	1.6534103	0.0986524	Lean R
Treatment - Control	6	3.612111	2.777914	775	1.3002961	0.1938861	Weak R
Treatment - Control	7	3.440814	3.459177	775	0.9946917	0.3201967	Strong R

Factor models for panel plots

predict_indiv_factor <- lm_robust(SPV_indiv_predict ~ condition_dummy * factor(party_numeric),
                                  data = data_clean, se_type = "HC2")
emm_predict_indiv_factor <- emmeans(predict_indiv_factor, ~ condition_dummy | party_numeric,
                                    vcov. = vcov(predict_indiv_factor))
contr_predict_indiv_factor_df <- confint(contrast(emm_predict_indiv_factor,
                                                  method = list("Treatment - Control" = c(-1, 1)),
                                                  by = "party_numeric")) |>
  as.data.frame() |>
  mutate(party_numeric = as.numeric(as.character(party_numeric)))

predict_meta_R_factor <- lm_robust(SPV_meta_predict ~ condition_dummy * factor(party_numeric),
                                   data = data_clean, se_type = "HC2")
emm_predict_meta_R_factor <- emmeans(predict_meta_R_factor, ~ condition_dummy | party_numeric,
                                     vcov. = vcov(predict_meta_R_factor))
contr_predict_meta_R_factor_df <- confint(contrast(emm_predict_meta_R_factor,
                                                   method = list("Treatment - Control" = c(-1, 1)),
                                                   by = "party_numeric")) |>
  as.data.frame() |>
  mutate(party_numeric = as.numeric(as.character(party_numeric)))

p_predict_indiv <- make_SPV_panels(
  "SPV_indiv_predict",
  contr_predict_indiv_partyaffil_df,
  contr_predict_indiv_factor_df,
  "Individual SPV: Predict\n(raw mean)",
  "Figure: If Opposing Party Wins Next Election — Individual Ratings",
  color_A = c("Control R" = "darkorchid", "Treatment R" = "plum",
              "Control D" = "orchid",      "Treatment D" = "thistle"))

p_predict_meta_R <- make_SPV_panels(
  "SPV_meta_predict",
  contr_predict_meta_R_partyaffil_df,
  contr_predict_meta_R_factor_df,
  "Meta-Perceptions of Republican SPV: Predict\n(raw mean)",
  "Figure: If Opposing Party Wins Next Election — Meta-Perceptions",
  color_A = c("Control R" = "darkorchid", "Treatment R" = "plum",
              "Control D" = "orchid",      "Treatment D" = "thistle"))

Binned Scenarios SPV

Secondarily to the scenario-specific models, I plan to combine the four scenario-specific questions and the two broad-based questions into two, distinct averages measuring individual and meta-perceptions of Republicans’ SPV. I will then apply the same Ordinary Least Squares (OLS) linear regression models with heteroskedasticity-robust standard errors to these aggregated variables, including the exact same covariates as for each of the six individual questions regarding SPV scenarios.

I deviated from the pre-analysis plan by removing the additional specified covariates from each model, as this simplifies the ability to draw conclusions about the average treatment effect. The only covariates included are PID_dummy, which represents the dichotomous effect of being a Democrat versus a Republican, and PID_bin, the grouped version of partisan self-concept.

Create four baseline models

lm_indiv_scenario <- lm_robust(SPV_indiv_scenario ~ condition_dummy, data = data_clean)
lm_indiv_broad <- lm_robust(SPV_indiv_broad ~ condition_dummy, data = data_clean)
lm_meta_R_scenario <- lm_robust(SPV_meta_scenario ~ condition_dummy, data = data_clean)
lm_meta_R_broad <- lm_robust(SPV_meta_broad ~ condition_dummy, data = data_clean)

term	estimate	std.error	statistic	p.value	conf.low	conf.high	df	outcome
(Intercept)	25.532	1.029	24.801	0.00	23.511	27.553	780	SPV_indiv_scenario
condition_dummy	2.373	1.525	1.556	0.12	-0.621	5.367	780	SPV_indiv_scenario

term	estimate	std.error	statistic	p.value	conf.low	conf.high	df	outcome
(Intercept)	13.268	1.089	12.183	0.000	11.130	15.406	784	SPV_indiv_broad
condition_dummy	2.017	1.619	1.246	0.213	-1.161	5.195	784	SPV_indiv_broad

term	estimate	std.error	statistic	p.value	conf.low	conf.high	df	outcome
(Intercept)	39.515	1.264	31.266	0.000	37.034	41.996	781	SPV_meta_scenario
condition_dummy	5.646	1.906	2.961	0.003	1.903	9.388	781	SPV_meta_scenario

term	estimate	std.error	statistic	p.value	conf.low	conf.high	df	outcome
(Intercept)	33.361	1.496	22.297	0.000	30.424	36.298	785	SPV_meta_broad
condition_dummy	5.446	2.242	2.429	0.015	1.045	9.847	785	SPV_meta_broad

INDIV_SCENARIO

# A: bivariate
lm_indiv_1 <- lm(SPV_indiv_scenario ~ condition_dummy, data = data_clean)
# B: multivariate
lm_indiv_1_cov <- lm(SPV_indiv_scenario ~ condition_dummy * PID_dummy * PID_bin,
                     data = data_clean)

term	estimate	std.error	statistic	p.value
(Intercept)	25.532	1.074	23.779	0.00
condition_dummy	2.373	1.524	1.557	0.12

term	estimate	std.error	statistic	p.value
(Intercept)	24.096	3.307	7.285	0.000
condition_dummy	10.277	4.731	2.172	0.030
PID_dummy	-19.442	5.549	-3.504	0.000
PID_bin	1.379	2.017	0.683	0.495
condition_dummy:PID_dummy	-3.594	8.072	-0.445	0.656
condition_dummy:PID_bin	-4.166	2.861	-1.457	0.146
PID_dummy:PID_bin	11.729	3.407	3.443	0.001
condition_dummy:PID_dummy:PID_bin	-0.921	5.019	-0.184	0.854

INDIV_BROAD

# A: bivariate
lm_indiv_2 <- lm(SPV_indiv_broad ~ condition_dummy, data = data_clean)
# B: multivariate
lm_indiv_2_cov <- lm(SPV_indiv_broad ~ condition_dummy * PID_dummy * PID_bin,
                     data = data_clean)

term	estimate	std.error	statistic	p.value
(Intercept)	13.268	1.141	11.626	0.000
condition_dummy	2.017	1.618	1.246	0.213

term	estimate	std.error	statistic	p.value
(Intercept)	11.304	3.612	3.129	0.002
condition_dummy	13.043	5.120	2.547	0.011
PID_dummy	-10.285	5.997	-1.715	0.087
PID_bin	1.547	2.196	0.704	0.481
condition_dummy:PID_dummy	-16.827	8.697	-1.935	0.053
condition_dummy:PID_bin	-6.082	3.090	-1.968	0.049
PID_dummy:PID_bin	6.296	3.665	1.718	0.086
condition_dummy:PID_dummy:PID_bin	7.303	5.395	1.354	0.176

META_R_SCENARIO

# A: bivariate
lm_meta_R_1 <- lm(SPV_meta_scenario ~ condition_dummy, data = data_clean)
# B: multivariate
lm_meta_R_1_cov <- lm(SPV_meta_scenario ~ condition_dummy * PID_dummy * PID_bin,
                      data = data_clean)

term	estimate	std.error	statistic	p.value
(Intercept)	39.515	1.339	29.519	0.000
condition_dummy	5.646	1.904	2.965	0.003

term	estimate	std.error	statistic	p.value
(Intercept)	42.586	4.158	10.243	0.000
condition_dummy	14.340	5.947	2.411	0.016
PID_dummy	-22.266	6.972	-3.194	0.001
PID_bin	-0.156	2.536	-0.061	0.951
condition_dummy:PID_dummy	7.954	10.150	0.784	0.433
condition_dummy:PID_bin	-5.355	3.594	-1.490	0.137
PID_dummy:PID_bin	9.710	4.271	2.274	0.023
condition_dummy:PID_dummy:PID_bin	-7.892	6.317	-1.249	0.212

META_R_BROAD

# A: bivariate
lm_meta_R_2 <- lm(SPV_meta_broad ~ condition_dummy, data = data_clean)
# B: multivariate
lm_meta_R_2_cov <- lm(SPV_meta_broad ~ condition_dummy * PID_dummy * PID_bin,
                     data = data_clean)

term	estimate	std.error	statistic	p.value
(Intercept)	33.361	1.579	21.126	0.000
condition_dummy	5.446	2.240	2.431	0.015

term	estimate	std.error	statistic	p.value
(Intercept)	39.973	4.761	8.396	0.000
condition_dummy	7.090	6.748	1.051	0.294
PID_dummy	-30.992	7.917	-3.915	0.000
PID_bin	0.859	2.898	0.297	0.767
condition_dummy:PID_dummy	7.535	11.485	0.656	0.512
condition_dummy:PID_bin	-1.731	4.078	-0.425	0.671
PID_dummy:PID_bin	6.548	4.840	1.353	0.177
condition_dummy:PID_dummy:PID_bin	-7.331	7.135	-1.027	0.305

Independent-Samples T-Tests

Lastly, in line with earlier work by Mernyk et al. (2022), I plan to perform three independent-samples t-tests comparing Republican to Democratic participants regarding their predicted and actual levels of support for political violence. The first t-test will compare Republicans’ reported individual SPV to Democrats’ meta-perceptions of Republicans’ SPV, while the second t-test will compare Republicans’ individual SPV to their own meta-perceptions of co-partisans’ SPV. The third t-test will compare meta-perceptions of Republicans’ SPV between Republicans and Democrats. I will use the standard p<.05 criteria in our two-tailed T-test to determine whether significant differences exist between individual support of partisan violence and meta-perceptions of Republicans’ support for partisan violence. For all t-tests, I plan to calculate the effect size as Cohen’s d.

# 8. outparty metaperceptions
# 8.1: actual views of R vs. how D think R will respond
t.test(na.omit((data_clean %>% filter(party_numeric >= 4))$SPV_individual),
       na.omit((data_clean %>% filter(party_numeric <= 3))$SPV_meta_outgroup))

## 
##  Welch Two Sample t-test
## 
## data:  na.omit((data_clean %>% filter(party_numeric >= 4))$SPV_individual) and na.omit((data_clean %>% filter(party_numeric <= 3))$SPV_meta_outgroup)
## t = -14.191, df = 628.49, p-value < 2.2e-16
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
##  -28.62956 -21.66938
## sample estimates:
## mean of x mean of y 
##  19.62387  44.77333

# 8.2: effect size
cohen.d(na.omit((data_clean %>% filter(party_numeric >= 4))$SPV_individual),
       na.omit((data_clean %>% filter(party_numeric <= 3))$SPV_meta_outgroup))

## 
## Cohen's d
## 
## d estimate: -0.996642 (large)
## 95 percent confidence interval:
##      lower      upper 
## -1.1540616 -0.8392224

# 9. inparty metaperceptions
# 9.1: actual views of R vs. how R think R will respond
t.test(na.omit((data_clean %>% filter(party_numeric >= 4))$SPV_individual),
       na.omit((data_clean %>% filter(party_numeric >= 4))$SPV_meta_ingroup))

## 
##  Welch Two Sample t-test
## 
## data:  na.omit((data_clean %>% filter(party_numeric >= 4))$SPV_individual) and na.omit((data_clean %>% filter(party_numeric >= 4))$SPV_meta_ingroup)
## t = -6.0073, df = 515, p-value = 3.568e-09
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
##  -14.997798  -7.605754
## sample estimates:
## mean of x mean of y 
##  19.62387  30.92564

# 9.2: effect size
cohen.d(na.omit((data_clean %>% filter(party_numeric >= 4))$SPV_individual),
        na.omit((data_clean %>% filter(party_numeric >= 4))$SPV_meta_ingroup))

## 
## Cohen's d
## 
## d estimate: -0.5283701 (medium)
## 95 percent confidence interval:
##      lower      upper 
## -0.7041666 -0.3525736

# 10. are democrats' inaccuracies different from republicans' inaccuracies?
# 10.1: how D think R will respond vs. how R think R will respond
t.test(na.omit((data_clean %>% filter(party_numeric <= 3))$SPV_meta_outgroup),
       na.omit((data_clean %>% filter(party_numeric >= 4))$SPV_meta_ingroup))

## 
##  Welch Two Sample t-test
## 
## data:  na.omit((data_clean %>% filter(party_numeric <= 3))$SPV_meta_outgroup) and na.omit((data_clean %>% filter(party_numeric >= 4))$SPV_meta_ingroup)
## t = 7.7772, df = 629.73, p-value = 3.04e-14
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
##  10.35117 17.34422
## sample estimates:
## mean of x mean of y 
##  44.77333  30.92564

# 10.2: effect size
cohen.d(na.omit((data_clean %>% filter(party_numeric <= 3))$SPV_meta_outgroup),
        na.omit((data_clean %>% filter(party_numeric >= 4))$SPV_meta_ingroup))

## 
## Cohen's d
## 
## d estimate: 0.547227 (medium)
## 95 percent confidence interval:
##     lower     upper 
## 0.3959149 0.6985390