I. Quick Summary
Results from Media Trust/Consumption difference/combined measures
Results from media perception difference score
Running the initial model predicting anger from a media difference score (Fox - Other) resulted in: a) No main effect of media difference score (p = .97); and b) A significant two-way interaction between partisanship (Dem vs. Rep) and media difference score (b = 0.58, p < .001).
Interaction was such that effect of Fox - Other was somewhat larger for Republicans (b = 0.34, p < .001) than for Democrats (b = -0.24, p < .001).
Results from media perception combined score
Running the initial model predicting anger from a combined score (Fox + Other) resulted in: a) No main effect of combined media measure (p = .54); and b) No interaction between partisanship (Dem vs. Rep) and media difference score (p = .61).
Results from only Media Consumption difference/combined measures
Results from media exposure difference score
Running the initial model predicting anger from a media difference score (Fox - Other) resulted in: a) No main effect of media difference score (p = .74); and b) A significant two-way interaction between partisanship (Dem vs. Rep) and media difference score (b = 0.46, p < .001).
Interaction was such that effect of Fox - Other was smaller for Democrats (b = -0.19, p = .009) than for Republicans (b = 0.27, p < .001).
Results from media exposure combined score
Running the initial model predicting anger from a combined score (Fox + Other) resulted in: a) Main effect of combined media measure (b = 0.09, p = .009); and b) No interaction between partisanship (Dem vs. Rep) and media difference score (p = .70).
II. Data & Codes
a. Data & libraries
b. Compose variables
c. Codes & indices
II. (1) Questions from Leaf on 2/10/2021
a. Regression estimating anger from partisanship and media exposure
b. Is media exposure associated with anger?
No
c. Do polarized emotions get smaller after accounting for media? (i.e., Is there an interaction between partisanship and media exposure on anger?)
There is an interaction between partisanship (Dem vs. Rep) and media exposure/trust (p < .001) as well as partisanship (Dem vs. Rep) and Fox exposure/trust (p < .001).
d. Graph
i. Anger by media perception and party
ii. Anger by Fox perception and party
(2) Mediation Analyses - Baron & Kenny approach
1. Y ~ X
Yes, p < .001
2. Z ~ X
Yes, p < .001
3. Y ~ Z + X
No different – b = -1.28 in the Y ~ X model, b = -1.23 in the model including the mediator
4. Mediation function!
III. (1) Follow-up questions from 2/12
Leaf’s comments: For media perception and anger, drop the fox*media interaction; we don’t have a great theoretical reason to include it.
The results are otherwise a little tricky to interpret. For OtherMedia, estimated at the average level of Fox (which is below group average for Rs and below group average for Ds), perception makes Ds more angry and Rs less angry. For Fox, the reverse pattern is true. This might reflect that anger increases with exposure to only one type of media sources: Fox and no other source, or every other source and not fox.
See what happens when instead using the difference (Fox - Others) and the sum (Fox + Others) as predictors.
a. Regression predicting anger from partisanship and media measures dropping fox*media interaction
anger2 <- lm(anger ~ (pDem_Rep + pInd_Not) * (foxPerception.c + mediaPerception.c), data = d)
summary(anger2)##
## Call:
## lm(formula = anger ~ (pDem_Rep + pInd_Not) * (foxPerception.c +
## mediaPerception.c), data = d)
##
## Residuals:
## Min 1Q Median 3Q Max
## -3.6608 -1.7090 -0.3631 1.4932 5.0886
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 2.89777 0.07368 39.329 < 2e-16 ***
## pDem_Rep 0.23044 0.16256 1.418 0.1566
## pInd_Not 0.37186 0.17095 2.175 0.0298 *
## foxPerception.c -0.01927 0.06774 -0.284 0.7761
## mediaPerception.c -0.04469 0.09084 -0.492 0.6229
## pDem_Rep:foxPerception.c 0.47882 0.10859 4.410 1.13e-05 ***
## pDem_Rep:mediaPerception.c -0.89428 0.17302 -5.169 2.76e-07 ***
## pInd_Not:foxPerception.c 0.12545 0.18147 0.691 0.4895
## pInd_Not:mediaPerception.c -0.25069 0.22894 -1.095 0.2737
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 2.021 on 1204 degrees of freedom
## (29 observations deleted due to missingness)
## Multiple R-squared: 0.05728, Adjusted R-squared: 0.05101
## F-statistic: 9.144 on 8 and 1204 DF, p-value: 2.845e-12No remarkable changes from fully interactive model (higher-order interactions weren’t significant in first go-around so this checks out).
b. Composing difference score and combined media measures
# difference score
d$diffFoxOther <- d$foxPerception - d$mediaPerception
# combined score
d$addFoxOther <- d$foxPerception + d$mediaPerceptionc. Regression w/ difference score
anger.diff <- lm(anger ~ (pDem_Rep + pInd_Not) * diffFoxOther, data = d)
summary(anger.diff)##
## Call:
## lm(formula = anger ~ (pDem_Rep + pInd_Not) * diffFoxOther, data = d)
##
## Residuals:
## Min 1Q Median 3Q Max
## -3.595 -1.738 -0.374 1.574 4.786
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 2.925713 0.072376 40.424 < 2e-16 ***
## pDem_Rep 0.318881 0.157444 2.025 0.0430 *
## pInd_Not 0.431508 0.169569 2.545 0.0111 *
## diffFoxOther -0.002481 0.062793 -0.040 0.9685
## pDem_Rep:diffFoxOther 0.583100 0.098547 5.917 4.27e-09 ***
## pInd_Not:diffFoxOther 0.159799 0.169196 0.944 0.3451
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 2.024 on 1207 degrees of freedom
## (29 observations deleted due to missingness)
## Multiple R-squared: 0.05182, Adjusted R-squared: 0.0479
## F-statistic: 13.19 on 5 and 1207 DF, p-value: 1.557e-12d. Regression w/ combined measure
anger.add <- lm(anger ~ (pDem_Rep + pInd_Not) * addFoxOther, data = d)
summary(anger.add)##
## Call:
## lm(formula = anger ~ (pDem_Rep + pInd_Not) * addFoxOther, data = d)
##
## Residuals:
## Min 1Q Median 3Q Max
## -2.6023 -1.6643 -0.5311 1.4715 4.3718
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 3.24675 0.21372 15.191 <2e-16 ***
## pDem_Rep 0.21502 0.43940 0.489 0.625
## pInd_Not 0.84250 0.51841 1.625 0.104
## addFoxOther -0.02659 0.04345 -0.612 0.541
## pDem_Rep:addFoxOther 0.04408 0.08582 0.514 0.608
## pInd_Not:addFoxOther -0.03162 0.10765 -0.294 0.769
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 2.055 on 1207 degrees of freedom
## (29 observations deleted due to missingness)
## Multiple R-squared: 0.02329, Adjusted R-squared: 0.01924
## F-statistic: 5.756 on 5 and 1207 DF, p-value: 2.935e-05e. Descriptives of media difference score and combined measure
# difference score by party
aggregate(d$diffFoxOther, list(d$party_factor), FUN = mean, na.rm = T)## Group.1 x
## 1 Democrat -0.9196957
## 2 Republican 0.9673349
## 3 Independent -0.1943609# combined score by party
aggregate(d$addFoxOther, list(d$party_factor), FUN = mean, na.rm = T)## Group.1 x
## 1 Democrat 4.853506
## 2 Republican 4.944303
## 3 Independent 4.510150f. Graphs
i. Anger by media difference and party
media.diff.plot <- ggplot(d[!is.na(d$party_factor),], aes(x = diffFoxOther, y = anger)) +
geom_smooth(method = 'lm', aes(x = diffFoxOther, y = anger, col = party_factor)) +
facet_wrap(~party_factor)
media.diff.plot + theme_classic() +
xlab("Difference between Fox and Other Media Trust/Exposure") +
ylab("Anger") +
theme(legend.position = c(1000,1000)) +
coord_cartesian(xlim = c(-4,4),ylim = c(1, 7)) +
scale_x_continuous(breaks = seq(-4,4,1)) +
scale_y_continuous(breaks = c(1,2,3,4,5,6,7)) +
scale_color_manual(values = c("dodgerblue","red3","grey42"))## `geom_smooth()` using formula 'y ~ x'## Warning: Removed 15 rows containing non-finite values (stat_smooth).
ii. Anger by Fox perception and party
media.add.plot <- ggplot(d[!is.na(d$party_factor),], aes(x = addFoxOther, y = anger)) +
geom_smooth(method = 'lm', aes(x = addFoxOther, y = anger, col = party_factor)) +
facet_wrap(~party_factor)
media.add.plot + theme_classic() +
xlab("Combined Fox and Other Media Trust/Exposure") +
ylab("Anger") +
theme(legend.position = c(1000,1000)) +
coord_cartesian(xlim = c(1,10),ylim = c(1, 7)) +
scale_x_continuous(breaks = seq(1,10,1)) +
scale_y_continuous(breaks = seq(1,7,1)) +
scale_color_manual(values = c("dodgerblue","red3","grey42"))## `geom_smooth()` using formula 'y ~ x'## Warning: Removed 15 rows containing non-finite values (stat_smooth).
(2) All analyses from III.1 run with just media consumption
mediaExposure_5 = Fox News mediaExposure = all but Fox News
a. Regression predicting anger from partisanship and media measures
anger3 <- lm(anger ~ (pDem_Rep + pInd_Not) * (mediaExposure_5 + mediaExposure), data = d)
summary(anger3)##
## Call:
## lm(formula = anger ~ (pDem_Rep + pInd_Not) * (mediaExposure_5 +
## mediaExposure), data = d)
##
## Residuals:
## Min 1Q Median 3Q Max
## -3.3571 -1.7731 -0.3797 1.5787 4.6790
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 2.68998 0.16036 16.775 < 2e-16 ***
## pDem_Rep 0.84381 0.34504 2.446 0.014607 *
## pInd_Not 0.81274 0.37842 2.148 0.031935 *
## mediaExposure_5 0.02564 0.06332 0.405 0.685585
## mediaExposure 0.12214 0.08391 1.456 0.145766
## pDem_Rep:mediaExposure_5 0.38391 0.10084 3.807 0.000148 ***
## pDem_Rep:mediaExposure -0.69667 0.15796 -4.411 1.12e-05 ***
## pInd_Not:mediaExposure_5 0.13835 0.16994 0.814 0.415739
## pInd_Not:mediaExposure -0.34249 0.21255 -1.611 0.107368
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 2.027 on 1205 degrees of freedom
## (28 observations deleted due to missingness)
## Multiple R-squared: 0.05076, Adjusted R-squared: 0.04446
## F-statistic: 8.055 on 8 and 1205 DF, p-value: 1.237e-10b. Composing difference score and combined media measures
# difference score
d$diffFoxOther.cons <- d$mediaExposure_5 - d$mediaExposure
# combined score
d$addFoxOther.cons <- d$mediaExposure_5 + d$mediaExposurec. Regression w/ difference score
anger.diff.cons <- lm(anger ~ (pDem_Rep + pInd_Not) * diffFoxOther.cons, data = d)
summary(anger.diff.cons)##
## Call:
## lm(formula = anger ~ (pDem_Rep + pInd_Not) * diffFoxOther.cons,
## data = d)
##
## Residuals:
## Min 1Q Median 3Q Max
## -3.3371 -1.7551 -0.3401 1.6566 4.6183
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 2.98309 0.06996 42.639 < 2e-16 ***
## pDem_Rep 0.20175 0.15005 1.345 0.17902
## pInd_Not 0.48000 0.16545 2.901 0.00378 **
## diffFoxOther.cons -0.01993 0.06049 -0.330 0.74181
## pDem_Rep:diffFoxOther.cons 0.46397 0.09478 4.895 1.11e-06 ***
## pInd_Not:diffFoxOther.cons 0.18678 0.16307 1.145 0.25227
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 2.032 on 1208 degrees of freedom
## (28 observations deleted due to missingness)
## Multiple R-squared: 0.04418, Adjusted R-squared: 0.04022
## F-statistic: 11.17 on 5 and 1208 DF, p-value: 1.543e-10d. Regression w/ combined measure
anger.add.cons <- lm(anger ~ (pDem_Rep + pInd_Not) * addFoxOther.cons, data = d)
summary(anger.add.cons)##
## Call:
## lm(formula = anger ~ (pDem_Rep + pInd_Not) * addFoxOther.cons,
## data = d)
##
## Residuals:
## Min 1Q Median 3Q Max
## -3.0941 -1.8479 -0.4305 1.6043 4.5331
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 2.73817 0.15556 17.602 < 2e-16 ***
## pDem_Rep 0.27945 0.32832 0.851 0.39485
## pInd_Not 0.75285 0.37157 2.026 0.04297 *
## addFoxOther.cons 0.09408 0.03600 2.613 0.00907 **
## pDem_Rep:addFoxOther.cons 0.02755 0.07085 0.389 0.69743
## pInd_Not:addFoxOther.cons -0.03357 0.08934 -0.376 0.70712
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 2.048 on 1208 degrees of freedom
## (28 observations deleted due to missingness)
## Multiple R-squared: 0.02879, Adjusted R-squared: 0.02477
## F-statistic: 7.161 on 5 and 1208 DF, p-value: 1.294e-06e. Descriptives of media difference score and combined measure
# difference score by party
aggregate(d$diffFoxOther.cons, list(d$party_factor), FUN = mean, na.rm = T)## Group.1 x
## 1 Democrat -0.46223509
## 2 Republican 1.15394089
## 3 Independent 0.02481203# combined score by party
aggregate(d$addFoxOther.cons, list(d$party_factor), FUN = mean, na.rm = T)## Group.1 x
## 1 Democrat 4.069378
## 2 Republican 4.462438
## 3 Independent 3.638346f. Graphs
i. Anger by media consumption difference and party
cons.diff.plot <- ggplot(d[!is.na(d$party_factor),], aes(x = diffFoxOther.cons, y = anger)) +
geom_smooth(method = 'lm', aes(x = diffFoxOther.cons, y = anger, col = party_factor)) +
facet_wrap(~party_factor)
cons.diff.plot + theme_classic() +
xlab("Difference between Fox and Other Media Exposure") +
ylab("Anger") +
theme(legend.position = c(1000,1000)) +
coord_cartesian(xlim = c(-4,4),ylim = c(1, 7)) +
scale_x_continuous(breaks = seq(-4,4,1)) +
scale_y_continuous(breaks = c(1,2,3,4,5,6,7)) +
scale_color_manual(values = c("dodgerblue","red3","grey42"))## `geom_smooth()` using formula 'y ~ x'## Warning: Removed 14 rows containing non-finite values (stat_smooth).
ii. Anger by combined media consumption and party
cons.add.plot <- ggplot(d[!is.na(d$party_factor),], aes(x = addFoxOther.cons, y = anger)) +
geom_smooth(method = 'lm', aes(x = addFoxOther.cons, y = anger, col = party_factor)) +
facet_wrap(~party_factor)
cons.add.plot + theme_classic() +
xlab("Combined Fox and Other Media Exposure") +
ylab("Anger") +
theme(legend.position = c(1000,1000)) +
coord_cartesian(xlim = c(1,10),ylim = c(1, 7)) +
scale_x_continuous(breaks = seq(1,10,1)) +
scale_y_continuous(breaks = seq(1,7,1)) +
scale_color_manual(values = c("dodgerblue","red3","grey42"))## `geom_smooth()` using formula 'y ~ x'## Warning: Removed 14 rows containing non-finite values (stat_smooth).