Are immigration policy preferences based on accurate stereotypes?
Emil O. W. Kirkegaard, Noah Carl, Julius D. Bjerrekær
Start
options(digits = 3)
options(contrasts=c("contr.treatment", "contr.treatment"))
library(pacman)
p_load(ggrepel, lavaanPlot, mediation, kirkegaard, dplyr, readr, googlesheets4, lavaan, polycor, rms, readxl, dkstat, ggeffects)
theme_set(theme_bw())
#ad hoc func
Inf_to_NA = function(x) {
x[is.infinite(x)] = NA
x
}
describe = function(x) psych::describe(x) %>% as.matrix()Data
#load data
silence({
kb16 = read_rds("data/Kirkegaard and Bjerrekær 2016 data.rds")
kb16_full = read_excel("data/Kirkegaard and Bjerrekær 2016 data raw.xlsx")
kb16_aggr = read_rds("data/Kirkegaard and Bjerrekær 2016 data aggr.rds") %>%
rownames_to_column("ISO")
dk_fiscal = read_csv("data/denmark.csv") %>%
mutate(
#add synonym to avoid breaking existing code
dnk_net_fiscal_contribution = dnk_net_fiscal_contribution_2014
)
#merge
dk_fiscal = full_join(dk_fiscal, kb16_aggr, by = "ISO") %>%
mutate(
dk_fiscal = dnk_net_fiscal_contribution_2014,
dk_benefits_use = Denmark.Social_benefits_30_39,
Muslim_frac = IslamPewResearch2010
) %>%
arrange(
ISO
)
#new data
#old google sheets package code
# gs = gs_url("https://docs.google.com/spreadsheets/d/1jsaXush5hD2rlyhzA0VaB8LR1eRb1h9mJOphTwCgeiU/edit#gid=1277242299")
# d = gs_read(gs)
#new package
# d = read_sheet("https://docs.google.com/spreadsheets/d/1jsaXush5hD2rlyhzA0VaB8LR1eRb1h9mJOphTwCgeiU/edit#gid=1277242299")
#local copy
# d = read_excel("data/this survey.xlsx")
d = read_excel("data/this survey 2020-03-01.xlsx")
#dst data
folk1a_2017q2 = read_excel("data/FOLK1A, 2017Q2.xlsx", range = "B3:D130") %>%
set_colnames(c("age", "men", "women")) %>%
mutate(age = age %>% str_replace(" år", "") %>% as.numeric(),
total = men + women)
HFUDD10_2017 = read_excel("data/HFUDD10 2017.xlsx", range = "B3:J12")
names(HFUDD10_2017)[1] = "Age"
})
#Extra Muslim data
#a few interpolated values for Muslim % for older countries
kirk17 = read_rds("data/Kirkegaard 2017.rds")
#UK data from Carl 2016
noah16 = readxl::read_xlsx("data/Immigrant Attitudes Crime.xlsx") %>%
#improve column names
df_legalize_names() %>%
#rename/transform to UK variants
mutate(
Net_opposition_UK = Net_opposition,
crime_rate_UK = c(standardize(Log_crime_rate) + standardize(Log_incarceration_rate)) %>% standardize(),
ISO = pu_translate(Country)
)## New names:
## * `` -> ...23
## * `` -> ...24Rename, recode, restructure
#controls
control_items = str_detect(names(d), "Vælg ") %>% which()
#subset, rename, score
d_controls = d[control_items]
d_controls[[5]] = d_controls[[5]] %in% 60:70
d_controls %<>% score_items(c("Samme antal indvandrere", "Færre indvandrere", "Negativt", "Positivt", T))
names(d_controls) = "control_" + 1:5
d_controls %<>% map_df(as.logical)
#prefs
pref_items = str_detect(names(d), "det samme antal indvandrere") %>% which() %>% setdiff(control_items)
econ_items = str_detect(names(d), "indvandrergrupper i Danmark bidrager til fælleskassen") %>% which() %>% setdiff(control_items)
#subset
d_prefs = d[pref_items]
d_econ = d[econ_items]
#sort variables by name function
df_sort_vars = function(x) {
x[sort(names(x))]
}
#rename, and to ISO-3
names(d_prefs) = str_match(names(d_prefs), "^[^:]+") %>% as.vector() %>% pu_translate()
names(d_econ) = str_match(names(d_econ), "^[^:]+") %>% as.vector() %>% pu_translate()
d_prefs %<>% df_sort_vars()
d_econ %<>% df_sort_vars()
#party agreements
party_items = str_detect(names(d), "med hvert politiske parti") %>% which() %>% setdiff(control_items)
d_parties = d[party_items]
names(d_parties) = str_match(names(d_parties), "^[^:]+")
#more rename
d$postcode = d$`URL Variable: pc` %>% as.numeric()
d$education = d$`Hvad er din højeste uddannelse?` %>%
mapvalues(c("Vil ikke oplyse", "0", "Anden uddannelse", "Folkeskole (til og med 10 klasse/real)"),
c(NA, "Grundskole", NA, "Grundskole"), warn_missing = F) %>%
ordered(levels = c("Grundskole",
"Gymnasial uddannelse (Alm. gymnasium/HF/HH/HGX/HTX)",
"Erhvervsuddannelser (HG, EFG, mesterlære m.v.)",
"Højere uddannelse af kort varighed (under 3 år)",
"Højere uddannelse af mellemlang varighed (3-4 år)",
"Højere uddannelse af lang varighed (5 år eller derover)"))
d$education_num = d$education %>% as.numeric()Quality control
#no variance in estimates
d_controls$control_no_var = map_lgl(1:nrow(d), function(x) {
d[x, econ_items] %>% unlist %>% all_the_same() %>% not()
})
#QC relations
polycor::hetcor(d_controls %>% as.data.frame())##
## Two-Step Estimates
##
## Correlations/Type of Correlation:
## control_1 control_2 control_3 control_4 control_5
## control_1 1 Polychoric Polychoric Polychoric Polychoric
## control_2 0.517 1 Polychoric Polychoric Polychoric
## control_3 0.621 0.709 1 Polychoric Polychoric
## control_4 0.693 0.759 0.91 1 Polychoric
## control_5 0.574 0.442 0.599 0.674 1
## control_no_var 0.226 0.417 0.547 0.604 0.215
## control_no_var
## control_1 Polychoric
## control_2 Polychoric
## control_3 Polychoric
## control_4 Polychoric
## control_5 Polychoric
## control_no_var 1
##
## Standard Errors:
## control_1 control_2 control_3 control_4 control_5
## control_1
## control_2 0.0609
## control_3 0.0504 0.0422
## control_4 0.0454 0.0383 0.0187
## control_5 0.0524 0.0587 0.0462 0.0423
## control_no_var 0.0846 0.0722 0.0601 0.0568 0.0751
##
## n = 850#QC scores
QC_score = rowSums(d_controls)
QC_score5 = rowSums(d_controls[-6])
#for each criterion
colMeans(d_controls)## control_1 control_2 control_3 control_4 control_5
## 0.868 0.854 0.798 0.826 0.744
## control_no_var
## 0.904#overall distribution of QC scores
table2(QC_score)#if we ignore the 6th
table2(QC_score5)#percent who was OK with first 5 who failed the 6th
table(QC_score5, d_controls$control_no_var)##
## QC_score5 FALSE TRUE
## 0 8 10
## 1 17 45
## 2 15 47
## 3 7 40
## 4 6 150
## 5 29 476#keep origs
origs = list(
d_econ = d_econ,
d_prefs = d_prefs,
d = d
)
#subset
d_econ = d_econ[QC_score == 6, ]
d_prefs = d_prefs[QC_score == 6, ]
d = d[QC_score == 6, ]Subset and more recoding
#recode
d_econ_num = d_econ %>% map_df(~plyr::mapvalues(., from = c("Meget negativt", "Negativt", "Lidt negativt", "Neutralt", "Lidt positivt", "Positivt", "Meget positivt"), to = 1:7, warn_missing = F) %>% as.numeric())
d_prefs_num = d_prefs %>% map_df(~plyr::mapvalues(., from = c("Ingen indvandrere", "Færre indvandrere", "Samme antal indvandrere", "Flere indvandrere"), to = c(1, 1, -1, -1), warn_missing = F) %>% as.numeric())
d_prefs_num_alt = d_prefs %>% map_df(~plyr::mapvalues(., from = c("Ingen indvandrere", "Færre indvandrere", "Samme antal indvandrere", "Flere indvandrere"), to = 1:4, warn_missing = F) %>% as.numeric())
#basic
d$gender = d$`URL Variable: gender` %>% plyr::mapvalues(c(1, 2), to = c("Male", "Female")) %>% factor(levels = c("Male", "Female"))
d$age = d$`URL Variable: yob` %>% subtract(2017) %>% abs()
#aux
d$Muslims_are_treated_well = d[[90]]
d$DK_admit_only_net_positive_immigrants = d[[91]]
d$nonwesterns_are_net_positive = d[[92]]
d$party_vote = d[[89]] %>%
mapvalues(from = c("Stemme blankt", "Vil ikke stemme"),
to = c("Vote blank", "Would not vote")) %>%
fct_relevel("Socialdemokraterne")
#binary vote split, to boost power
d$block_vote = d$party_vote %>%
as.character() %>%
mapvalues(
from = c("Socialdemokraterne", "Alternativet", "Enhedslisten", "Radikale Venstre", "Socialistisk Folkeparti", "Dansk Folkeparti", "Konservative Folkeparti", "Kristendemokraterne", "Liberal Alliance", "Nye Borgerlige", "Venstre"),
to = c(rep("left-block", 5), rep("right-block", 6))
) %>% factor() %>% fct_relevel("right-block")
d$block_vote %>% table2()#parties
d_parties = d[str_detect(names(d), "hvor 0 helt uenig, og 100 er helt enig") %>% which() %>% setdiff(control_items)]
names(d_parties) = str_match(names(d_parties), "[^:]+") %>% as.vector() %>% str_replace_all(" ", "_")
d = cbind(d, d_parties)
#recode path
d$path_parsed = plyr::alply(d$Path, .margins = 1, function(x) {
orders = c(A = str_locate(x, "9")[1], B = str_locate(x, "10")[1], C = str_locate(x, "11")[1])
str_c(names(sort(orders)), collapse = "")
}) %>% unlist()
#recode collection date
d$collected = d$`Date Submitted` %>% parse_date(format = "%d %B %Y %H:%M:%S")
#better country names
dk_fiscal$Names = pu_translate(dk_fiscal$ISO, reverse = T)
#add Muslim estimates for a few ex-countries
dk_fiscal[, c("Names", "Muslim_frac")] %>% .[miss_by_case(.) == 1, ]#join
dk_fiscal = left_join(dk_fiscal, kirk17[c("ISO", "Muslim")], by = "ISO")
#assert no missing
dk_fiscal[, c("Names", "Muslim")] %>% .[miss_by_case(.) == 1, ] %>% nrow() %equals% 0 %>% assert_that()## [1] TRUE#subset of countries included in the new survey
dk_fiscal_sub = dk_fiscal %>% filter(!is.na(dk_fiscal))
#exclude Syria (recode to NA)
assert_that(dk_fiscal_sub$ISO[26] == "SYR")## [1] TRUE#function to do it on the fly
no_SYR = function(x) {
if (is.data.frame(x)) {
x[26, ] = NA
return(x)
}
x[26] = NA
x
}
#sort dk_fiscal
dk_fiscal_sub = dplyr::arrange(dk_fiscal_sub, ISO)
#assert same order
assert_that(names(d_econ) %equals% names(d_prefs))## [1] TRUEassert_that(names(d_econ) %equals% dk_fiscal_sub$ISO)## [1] TRUE#aggregate
dk_fiscal_sub$mean_estimate_fiscal = colMeans(d_econ_num)
#aggregate prefs
dk_fiscal_sub$net_opposition = colMeans(d_prefs_num)
dk_fiscal_sub$mean_pref = colMeans(d_prefs_num_alt)
#split half to partially avoid the method variance
set.seed(1)
split = sample(c(T, F), size = nrow(d), replace = T)
dk_fiscal_sub$net_opposition_split1 = colMeans(d_prefs_num[split, ])
dk_fiscal_sub$net_opposition_split2 = colMeans(d_prefs_num[!split, ])
dk_fiscal_sub$mean_estimate_fiscal_split1 = colMeans(d_econ_num[split, ])
dk_fiscal_sub$mean_estimate_fiscal_split2 = colMeans(d_econ_num[!split, ])Reanalysis of Kirkegaard & Bjerrekær 2016
Determine the aggregate accuracy in the subset of 32 countries used for the present study.
#correlation between outcomes
GG_scatter(dk_fiscal_sub, "dk_benefits_use", "dk_fiscal", case_names = "Names") +
xlab("Benefits use in Denmark") +
ylab("Net fiscal contribution in Denmark")## `geom_smooth()` using formula 'y ~ x'
GG_scatter(dk_fiscal_sub %>% no_SYR(), "dk_benefits_use", "dk_fiscal", case_names = "Names") +
xlab("Benefits use in Denmark") +
ylab("Net fiscal contribution in Denmark") +
labs(caption = "Syria excluded")## `geom_smooth()` using formula 'y ~ x'
#stereotype accuracy in reduced sample
GG_scatter(dk_fiscal, "dk_benefits_use", "mean_estimate", case_names = "Names") +
xlab("Benefits use in Denmark") +
ylab("Mean estimate of benefits use in Denmark")## `geom_smooth()` using formula 'y ~ x'
GG_scatter(dk_fiscal_sub, "dk_benefits_use", "mean_estimate", case_names = "Names") +
xlab("Benefits use in Denmark") +
ylab("Mean estimate of benefits use in Denmark")## `geom_smooth()` using formula 'y ~ x'
#cross-accuracy
GG_scatter(dk_fiscal, "dk_fiscal", "mean_estimate", case_names = "Names") +
xlab("Net fiscal contribution") +
ylab("Mean estimate of benefits use in Denmark")## `geom_smooth()` using formula 'y ~ x'
GG_scatter(dk_fiscal %>% filter(ISO != "SYR"), "dk_fiscal", "mean_estimate", case_names = "Names") +
xlab("Net fiscal contribution") +
ylab("Mean estimate of benefits use in Denmark") +
labs(caption = "Syria excluded")## `geom_smooth()` using formula 'y ~ x'
#compare Muslim bias measures
#can we make a bias measure that works without ratio scale data?
#estimtes df
kb16_ests = kb16 %>% select(Afghanistan:Østrig)
#to ISO
colnames(kb16_ests) = pu_translate(colnames(kb16_ests))## No exact match: Bosnien Hercegovina## Best fuzzy match found: Bosnien Hercegovina -> Bosnien-Hercegovina with distance 1.00#reorder cols
kb16_ests = kb16_ests[sort(colnames(kb16_ests))]
#assert the order is the same
assert_that(all(colnames(kb16_ests) == dk_fiscal$ISO))## [1] TRUE#calculate new metrics
muslim_bias_scalefree = plyr::adply(kb16_ests, .margins = 1, function(r) {
d = tibble(
#names
name = dk_fiscal$Names,
#extract estimates
ests = r %>% unlist() %>% as.vector(),
#standardize within person
#this prevents elevation effects
ests_z = ests %>% standardize(),
#get outcomes
true_values = dk_fiscal$dk_benefits_use,
true_values_z = true_values %>% standardize(),
#Muslim, frac
muslim_pct = dk_fiscal$Muslim
)
#regress
fit = lm(true_values_z ~ ests_z, data = d)
#add more vars
d %<>% mutate(
#save resids, standardized
resids_z = fit %>% resid() %>% standardize()
)
# browser()
#return
data.frame(
#resid z x muslim
muslim_resid_r = wtd.cors(d$resids_z, d$muslim_pct) %>% as.vector() %>% `*`(-1),
#note that abs must be taken from the resids
muslim_resid_r_abs = wtd.cors(abs(d$resids_z), d$muslim_pct) %>% as.vector() %>% `*`(-1),
#mean methods
wtd_mean_resid_muslim = wtd_mean(d$resids_z, d$muslim_pct) %>% `*`(-1),
wtd_mean_resid_muslim_abs = wtd_mean(abs(d$resids_z), d$muslim_pct) %>% `*`(-1)
)
}, .expand = F)
#add to main data
kb16 = cbind(
kb16,
muslim_bias_scalefree[-1]
)
#person-level Muslim bias metrics
muslim_bias_vars = c("muslim_resid_r",
"wtd_mean_resid_muslim",
"muslim_elevation_error",
"muslim_error_r",
#abs versions
"muslim_resid_r_abs",
"wtd_mean_resid_muslim_abs",
"muslim_elevation_error_abs",
"muslim_error_abs_r"
)
#correlations
wtd.cors(kb16[muslim_bias_vars])## muslim_resid_r wtd_mean_resid_muslim
## muslim_resid_r 1.0000 1.0000
## wtd_mean_resid_muslim 1.0000 1.0000
## muslim_elevation_error 0.7105 0.7105
## muslim_error_r 0.6450 0.6450
## muslim_resid_r_abs 0.7508 0.7508
## wtd_mean_resid_muslim_abs 0.7802 0.7802
## muslim_elevation_error_abs -0.0673 -0.0673
## muslim_error_abs_r -0.0890 -0.0890
## muslim_elevation_error muslim_error_r
## muslim_resid_r 0.710 0.645
## wtd_mean_resid_muslim 0.710 0.645
## muslim_elevation_error 1.000 0.964
## muslim_error_r 0.964 1.000
## muslim_resid_r_abs 0.497 0.442
## wtd_mean_resid_muslim_abs 0.523 0.466
## muslim_elevation_error_abs -0.101 -0.262
## muslim_error_abs_r -0.274 -0.427
## muslim_resid_r_abs wtd_mean_resid_muslim_abs
## muslim_resid_r 0.7508 0.7802
## wtd_mean_resid_muslim 0.7508 0.7802
## muslim_elevation_error 0.4970 0.5228
## muslim_error_r 0.4417 0.4660
## muslim_resid_r_abs 1.0000 0.9896
## wtd_mean_resid_muslim_abs 0.9896 1.0000
## muslim_elevation_error_abs -0.0702 -0.0841
## muslim_error_abs_r -0.0880 -0.1018
## muslim_elevation_error_abs muslim_error_abs_r
## muslim_resid_r -0.0673 -0.089
## wtd_mean_resid_muslim -0.0673 -0.089
## muslim_elevation_error -0.1009 -0.274
## muslim_error_r -0.2625 -0.427
## muslim_resid_r_abs -0.0702 -0.088
## wtd_mean_resid_muslim_abs -0.0841 -0.102
## muslim_elevation_error_abs 1.0000 0.948
## muslim_error_abs_r 0.9480 1.000#subset
muslim_bias_vars2 = c("wtd_mean_resid_muslim",
"muslim_error_r",
"wtd_mean_resid_muslim_abs",
"muslim_error_abs_r")
#useful predictors from previous study still useful?
muslim_bias_useful_preds = c("Nationalism", "Conservatism", "Muslim_problem", "Future_perf_immi")
cor_matrix(kb16[c(muslim_bias_vars2, muslim_bias_useful_preds)], CI = .95) %>%
.[muslim_bias_vars2, muslim_bias_useful_preds]## Nationalism Conservatism
## wtd_mean_resid_muslim "0.12 [0.04 0.21]" "0.15 [0.06 0.23]"
## muslim_error_r "0.30 [0.21 0.38]" "0.18 [0.09 0.27]"
## wtd_mean_resid_muslim_abs "0.07 [-0.02 0.16]" "0.13 [0.04 0.22]"
## muslim_error_abs_r "-0.12 [-0.21 -0.03]" "-0.08 [-0.17 0.01]"
## Muslim_problem Future_perf_immi
## wtd_mean_resid_muslim "0.14 [0.05 0.23]" "-0.14 [-0.22 -0.05]"
## muslim_error_r "0.32 [0.24 0.40]" "-0.31 [-0.39 -0.23]"
## wtd_mean_resid_muslim_abs "0.09 [0.00 0.18]" "-0.13 [-0.22 -0.04]"
## muslim_error_abs_r "-0.14 [-0.23 -0.05]" "0.13 [0.04 0.21]"#sample size for cor matrix
kb16 %>% nrow()## [1] 484Aggregate
Descriptive
#plot contributions by origin
ggplot(dk_fiscal_sub, aes(fct_reorder(Names, dk_fiscal), dk_fiscal * 1000 / 7.45, fill = Muslim)) +
geom_bar(stat = "identity") +
scale_x_discrete("Country of origin") +
scale_y_continuous("Net fiscal contribution, 2014, in EUR/person/year", labels = scales::comma) +
scale_fill_gradient("Muslim%", label = scales::percent, low = "darkblue", high = "green", limits = c(0, 1)) +
theme(axis.text.x = element_text(angle = -30, hjust = 0))
GG_save("figures/fiscal_estimates.png")Accuracy
#test-retest of aggr stereotypes
GG_scatter(dk_fiscal_sub, "mean_estimate_fiscal", "mean_estimate", case_names = "Names", repel_names = T) +
xlab("Estimate of net fiscal contribution") +
ylab("Estimate of % of 30-39 year olds\non unemployment benefit")## `geom_smooth()` using formula 'y ~ x'
GG_save("figures/aggr_retest_stereotypes.png")## `geom_smooth()` using formula 'y ~ x'#without SYR
GG_scatter(dk_fiscal_sub %>% no_SYR(), "mean_estimate_fiscal", "mean_estimate", case_names = "Names", repel_names = T) +
xlab("Estimate of net fiscal contribution") +
ylab("Estimate of % of 30-39 year olds\non unemployment benefit") +
labs(caption = "Syria excluded")## `geom_smooth()` using formula 'y ~ x'
GG_save("figures/aggr_retest_stereotypes_noSYR.png")## `geom_smooth()` using formula 'y ~ x'#stereotype accuracy aggregate
GG_scatter(dk_fiscal_sub, "dk_fiscal", "mean_estimate_fiscal", case_names = "Names", repel_names = T) +
scale_y_continuous("Mean estimate of net fiscal contribution of group") +
scale_x_continuous("Group's net fiscal contribution (EUR/person/year)\nDanish Ministry of Finance estimate")## `geom_smooth()` using formula 'y ~ x'
GG_save("figures/aggregate_accuracy.png")## `geom_smooth()` using formula 'y ~ x'#without SYR
dk_fiscal_sub %>% no_SYR() %>%
GG_scatter("dk_fiscal", "mean_estimate_fiscal", case_names = "Names", repel_names = T) +
scale_y_continuous("Mean estimate of net fiscal contribution of group") +
scale_x_continuous("Group's net fiscal contribution\nDanish Ministry of Finance estimate") +
labs(caption = "Syria excluded")## `geom_smooth()` using formula 'y ~ x'
GG_save("figures/aggregate_accuracy_no_SYR.png")## `geom_smooth()` using formula 'y ~ x'#test
cor.test(dk_fiscal_sub$dnk_net_fiscal_contribution_2014, dk_fiscal_sub$mean_estimate_fiscal)##
## Pearson's product-moment correlation
##
## data: dk_fiscal_sub$dnk_net_fiscal_contribution_2014 and dk_fiscal_sub$mean_estimate_fiscal
## t = 7, df = 30, p-value = 3e-08
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## 0.637 0.902
## sample estimates:
## cor
## 0.806cor.test(dk_fiscal_sub$dnk_net_fiscal_contribution_2015, dk_fiscal_sub$mean_estimate_fiscal)##
## Pearson's product-moment correlation
##
## data: dk_fiscal_sub$dnk_net_fiscal_contribution_2015 and dk_fiscal_sub$mean_estimate_fiscal
## t = 9, df = 30, p-value = 2e-09
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## 0.698 0.920
## sample estimates:
## cor
## 0.842Muslim bias
#model
(aggregate_accuracy_fit = lm(dk_fiscal ~ mean_estimate_fiscal, data = dk_fiscal_sub, na.action = na.exclude) %>% MOD_summary())##
## ---- Model summary ----
## Model coefficients
## Beta SE CI_lower CI_upper
## mean_estimate_fiscal 0.806 0.108 0.586 1.03
##
##
## Model meta-data
## outcome N df R2 R2-adj. R2-cv
## 1 dk_fiscal 32 30 0.65 0.639 NA
##
##
## Etas from analysis of variance
## Eta Eta_partial
## mean_estimate_fiscal 0.806 0.806(aggregate_accuracy_fit_noSYR = lm(dk_fiscal ~ mean_estimate_fiscal, data = dk_fiscal_sub %>% no_SYR(), na.action = na.exclude) %>% MOD_summary())##
## ---- Model summary ----
## Model coefficients
## Beta SE CI_lower CI_upper
## mean_estimate_fiscal 0.847 0.0986 0.645 1.05
##
##
## Model meta-data
## outcome N df R2 R2-adj. R2-cv
## 1 dk_fiscal 31 29 0.718 0.708 NA
##
##
## Etas from analysis of variance
## Eta Eta_partial
## mean_estimate_fiscal 0.847 0.847#slopes
ols(dk_fiscal ~ mean_estimate_fiscal, data = dk_fiscal_sub)## Linear Regression Model
##
## ols(formula = dk_fiscal ~ mean_estimate_fiscal, data = dk_fiscal_sub)
##
## Model Likelihood Discrimination
## Ratio Test Indexes
## Obs 32 LR chi2 33.62 R2 0.650
## sigma47.4278 d.f. 1 R2 adj 0.639
## d.f. 30 Pr(> chi2) 0.0000 g 73.819
##
## Residuals
##
## Min 1Q Median 3Q Max
## -177.478 -13.408 5.013 21.852 94.585
##
##
## Coef S.E. t Pr(>|t|)
## Intercept -241.9446 31.0867 -7.78 <0.0001
## mean_estimate_fiscal 58.1153 7.7817 7.47 <0.0001
## ols(dk_fiscal ~ mean_estimate_fiscal, data = dk_fiscal_sub %>% no_SYR())## Frequencies of Missing Values Due to Each Variable
## dk_fiscal mean_estimate_fiscal
## 1 1
##
## Linear Regression Model
##
## ols(formula = dk_fiscal ~ mean_estimate_fiscal, data = dk_fiscal_sub %>%
## no_SYR())
##
##
## Model Likelihood Discrimination
## Ratio Test Indexes
## Obs 31 LR chi2 39.22 R2 0.718
## sigma33.2572 d.f. 1 R2 adj 0.708
## d.f. 29 Pr(> chi2) 0.0000 g 60.452
##
## Residuals
##
## Min 1Q Median 3Q Max
## -55.357 -23.245 -2.337 20.048 76.975
##
##
## Coef S.E. t Pr(>|t|)
## Intercept -200.2897 23.0083 -8.71 <0.0001
## mean_estimate_fiscal 48.9076 5.6942 8.59 <0.0001
## #save resids
dk_fiscal_sub$aggr_model_resid = resid(aggregate_accuracy_fit$model_obj) %>% standardize()
dk_fiscal_sub$aggr_model_resid_noSYR = resid(aggregate_accuracy_fit_noSYR$model_obj) %>% standardize()
#plot vs Muslim
dk_fiscal_sub %>%
GG_scatter("Muslim", "aggr_model_resid", case_names = "Names", repel_names = T) +
scale_x_continuous("Muslim population (%)", labels = scales::percent) +
scale_y_continuous("Stereotype estimation error\nNet fiscal contribution")## `geom_smooth()` using formula 'y ~ x'
GG_save("figures/aggregate_muslim_bias.png")## `geom_smooth()` using formula 'y ~ x'dk_fiscal_sub %>%
GG_scatter("Muslim", "aggr_model_resid_noSYR", case_names = "Names", repel_names = T) +
scale_x_continuous("Muslim population (%)", labels = scales::percent) +
scale_y_continuous("Stereotype estimation error\nNet fiscal contribution") +
labs(caption = "Syria excluded")## `geom_smooth()` using formula 'y ~ x'
GG_save("figures/aggregate_muslim_bias_noSYR.png")## `geom_smooth()` using formula 'y ~ x'#old data model
(aggregate_accuracy_fit_old = lm(dk_benefits_use ~ mean_estimate, data = dk_fiscal_sub, na.action = na.exclude) %>% MOD_summary())##
## ---- Model summary ----
## Model coefficients
## Beta SE CI_lower CI_upper
## mean_estimate 0.84 0.099 0.638 1.04
##
##
## Model meta-data
## outcome N df R2 R2-adj. R2-cv
## 1 dk_benefits_use 32 30 0.706 0.696 NA
##
##
## Etas from analysis of variance
## Eta Eta_partial
## mean_estimate 0.84 0.84(aggregate_accuracy_fit_old_all = lm(dk_benefits_use ~ mean_estimate, data = dk_fiscal, na.action = na.exclude) %>% MOD_summary())##
## ---- Model summary ----
## Model coefficients
## Beta SE CI_lower CI_upper
## mean_estimate 0.696 0.0871 0.522 0.87
##
##
## Model meta-data
## outcome N df R2 R2-adj. R2-cv
## 1 dk_benefits_use 70 68 0.484 0.477 NA
##
##
## Etas from analysis of variance
## Eta Eta_partial
## mean_estimate 0.696 0.696#save resids
#multiply by -1 because the outcome is reverse in desirability
dk_fiscal_sub$aggr_model_resid_old = resid(aggregate_accuracy_fit_old$model_obj) %>% standardize() %>% `*`(-1)
dk_fiscal$aggr_model_resid_old = resid(aggregate_accuracy_fit_old_all$model_obj) %>% standardize() %>% `*`(-1)
#old data, same countries, new metric
dk_fiscal_sub %>%
GG_scatter("Muslim", "aggr_model_resid_old", case_names = "Names", repel_names = T, text_pos = "tr") +
scale_x_continuous("Muslim population (%)", labels = scales::percent) +
scale_y_continuous("Stereotype residual")## `geom_smooth()` using formula 'y ~ x'
GG_save("figures/aggregate_muslim_bias_old_data_new_metric.png")## `geom_smooth()` using formula 'y ~ x'#old data, all countries, new metric
dk_fiscal %>%
GG_scatter("Muslim", "aggr_model_resid_old", case_names = "Names", repel_names = T) +
scale_x_continuous("Muslim population (%)", labels = scales::percent) +
scale_y_continuous("Stereotype residual")## `geom_smooth()` using formula 'y ~ x'
GG_save("figures/aggregate_muslim_bias_old_data_new_metric_all_countries.png")## `geom_smooth()` using formula 'y ~ x'#old data, same countries, old metric
dk_fiscal_sub %>%
GG_scatter("Muslim", "delta_est", case_names = "Names", repel_names = T) +
scale_x_continuous("Muslim population (%)", labels = scales::percent) +
scale_y_continuous("Stereotype residual")## `geom_smooth()` using formula 'y ~ x'
GG_save("figures/aggregate_muslim_bias_old_data.png")## `geom_smooth()` using formula 'y ~ x'#old data, all countries, old metric
dk_fiscal %>%
GG_scatter("Muslim", "delta_est", case_names = "Names", repel_names = T) +
scale_x_continuous("Muslim population (%)", labels = scales::percent) +
scale_y_continuous("Stereotype residual")## `geom_smooth()` using formula 'y ~ x'
GG_save("figures/aggregate_muslim_bias_old_data_all.png")## `geom_smooth()` using formula 'y ~ x'#cors
(subset_cors = dk_fiscal_sub %>% {cor_matrix(.[c("Muslim", "aggr_model_resid", "aggr_model_resid_old", "delta_est", "aggr_model_resid_noSYR")], CI = .95)})## Muslim aggr_model_resid
## Muslim "1" "-0.25 [-0.55 0.11]"
## aggr_model_resid "-0.25 [-0.55 0.11]" "1"
## aggr_model_resid_old "-0.11 [-0.44 0.25]" "0.62 [0.34 0.79]"
## delta_est "-0.39 [-0.65 -0.05]" "0.65 [0.40 0.82]"
## aggr_model_resid_noSYR "-0.30 [-0.59 0.06]" "0.96 [0.91 0.98]"
## aggr_model_resid_old delta_est
## Muslim "-0.11 [-0.44 0.25]" "-0.39 [-0.65 -0.05]"
## aggr_model_resid "0.62 [0.34 0.79]" "0.65 [0.40 0.82]"
## aggr_model_resid_old "1" "0.94 [0.88 0.97]"
## delta_est "0.94 [0.88 0.97]" "1"
## aggr_model_resid_noSYR "0.80 [0.63 0.90]" "0.82 [0.65 0.91]"
## aggr_model_resid_noSYR
## Muslim "-0.30 [-0.59 0.06]"
## aggr_model_resid "0.96 [0.91 0.98]"
## aggr_model_resid_old "0.80 [0.63 0.90]"
## delta_est "0.82 [0.65 0.91]"
## aggr_model_resid_noSYR "1"(all_cors = dk_fiscal %>% {cor_matrix(.[c("Muslim", "aggr_model_resid_old", "delta_est")], CI = .95)})## Muslim aggr_model_resid_old
## Muslim "1" "-0.27 [-0.48 -0.04]"
## aggr_model_resid_old "-0.27 [-0.48 -0.04]" "1"
## delta_est "-0.34 [-0.53 -0.12]" "0.99 [0.99 1.00]"
## delta_est
## Muslim "-0.34 [-0.53 -0.12]"
## aggr_model_resid_old "0.99 [0.99 1.00]"
## delta_est "1"#table
tibble(
data = c("new", "new", "old", "old", "old", "old"),
n = c(32, 31, 32, 70, 32, 70),
metric = c("Muslim resid r", "Muslim resid r", "Muslim resid r", "Muslim resid r", "Muslim error r", "Muslim error r"),
value = c(
subset_cors[1, 2],
subset_cors[1, 5],
subset_cors[1, 3],
all_cors[1, 2],
subset_cors[1, 4],
all_cors[1, 3]
)
)Immigrant preferences and stereotypes
#fiscal x prefs
GG_scatter(dk_fiscal_sub, "dk_fiscal", "net_opposition", case_names = "Names", repel_names = T) +
scale_x_continuous("Net fiscal contribution") +
scale_y_continuous("Net opposition")## `geom_smooth()` using formula 'y ~ x'
GG_save("figures/aggr_fiscal_net_opposition.png")## `geom_smooth()` using formula 'y ~ x'dk_fiscal_sub %>%
filter(ISO != "SYR") %>%
GG_scatter("dk_fiscal", "net_opposition", case_names = "Names", repel_names = T) +
scale_x_continuous("Net fiscal contribution") +
scale_y_continuous("Net opposition") +
labs(caption = "Syria excluded")## `geom_smooth()` using formula 'y ~ x'
GG_save("figures/aggr_fiscal_net_opposition_no_SYR.png")## `geom_smooth()` using formula 'y ~ x'#stereotypes x prefs
GG_scatter(dk_fiscal_sub, "mean_estimate_fiscal", "net_opposition", case_names = "Names", repel_names = T) +
scale_x_continuous("Mean estimate of net fiscal contribution") +
scale_y_continuous("Net opposition")## `geom_smooth()` using formula 'y ~ x'
GG_save("figures/aggr_stereotype_net_opposition.png")## `geom_smooth()` using formula 'y ~ x'dk_fiscal_sub %>%
filter(ISO != "SYR") %>%
GG_scatter("mean_estimate_fiscal", "net_opposition", case_names = "Names", repel_names = T) +
scale_x_continuous("Mean estimate of net fiscal contribution") +
scale_y_continuous("Net opposition") +
labs(caption = "Syria excluded")## `geom_smooth()` using formula 'y ~ x'
GG_save("figures/aggr_stereotype_net_opposition_no_SYR.png")## `geom_smooth()` using formula 'y ~ x'#path model
path_model = "net_opposition ~ mean_estimate_fiscal + dk_fiscal + Muslim
mean_estimate_fiscal ~ dk_fiscal
Muslim ~~ dk_fiscal"
(path_fit = lavaan::sem(path_model, data = dk_fiscal_sub, std.ov = T))## lavaan 0.6-5 ended normally after 45 iterations
##
## Estimator ML
## Optimization method NLMINB
## Number of free parameters 9
##
## Number of observations 32
##
## Model Test User Model:
##
## Test statistic 3.741
## Degrees of freedom 1
## P-value (Chi-square) 0.053(path_fit_noSYR = lavaan::sem(path_model, data = dk_fiscal_sub %>% no_SYR(), std.ov = T))## lavaan 0.6-5 ended normally after 39 iterations
##
## Estimator ML
## Optimization method NLMINB
## Number of free parameters 9
##
## Used Total
## Number of observations 31 32
##
## Model Test User Model:
##
## Test statistic 1.029
## Degrees of freedom 1
## P-value (Chi-square) 0.310fitmeasures(path_fit)## npar fmin chisq df
## 9.000 0.058 3.741 1.000
## pvalue baseline.chisq baseline.df baseline.pvalue
## 0.053 170.438 6.000 0.000
## cfi tli nnfi rfi
## 0.983 0.900 0.900 0.868
## nfi pnfi ifi rni
## 0.978 0.163 0.984 0.983
## logl unrestricted.logl aic bic
## -96.244 -94.373 210.487 223.679
## ntotal bic2 rmsea rmsea.ci.lower
## 32.000 195.622 0.293 0.000
## rmsea.ci.upper rmsea.pvalue rmr rmr_nomean
## 0.633 0.063 0.060 0.060
## srmr srmr_bentler srmr_bentler_nomean crmr
## 0.062 0.062 0.062 0.079
## crmr_nomean srmr_mplus srmr_mplus_nomean cn_05
## 0.079 0.061 0.061 33.861
## cn_01 gfi agfi pgfi
## 57.757 0.948 0.477 0.095
## mfi ecvi
## 0.958 0.679fitmeasures(path_fit_noSYR)## npar fmin chisq df
## 9.000 0.017 1.029 1.000
## pvalue baseline.chisq baseline.df baseline.pvalue
## 0.310 170.111 6.000 0.000
## cfi tli nnfi rfi
## 1.000 0.999 0.999 0.964
## nfi pnfi ifi rni
## 0.994 0.166 1.000 1.000
## logl unrestricted.logl aic bic
## -89.375 -88.860 196.750 209.655
## ntotal bic2 rmsea rmsea.ci.lower
## 31.000 181.616 0.031 0.000
## rmsea.ci.upper rmsea.pvalue rmr rmr_nomean
## 0.477 0.329 0.028 0.028
## srmr srmr_bentler srmr_bentler_nomean crmr
## 0.029 0.029 0.029 0.037
## crmr_nomean srmr_mplus srmr_mplus_nomean cn_05
## 0.037 0.029 0.029 116.691
## cn_01 gfi agfi pgfi
## 200.819 0.984 0.839 0.098
## mfi ecvi
## 1.000 0.614parameterestimates(path_fit, standardized = T)parameterestimates(path_fit_noSYR, standardized = T)#plot
lavaanPlot(model = path_fit, coefs = T, stand = T)lavaanPlot(model = path_fit_noSYR, coefs = T, stand = T)#simple modeling
MOD_summary(lm(net_opposition ~ mean_estimate_fiscal + dk_fiscal + Muslim, data = dk_fiscal_sub))##
## ---- Model summary ----
## Model coefficients
## Beta SE CI_lower CI_upper
## mean_estimate_fiscal -1.0607 0.0619 -1.18754 -0.934
## dk_fiscal 0.1265 0.0626 -0.00167 0.255
## Muslim 0.0304 0.0535 -0.07927 0.140
##
##
## Model meta-data
## outcome N df R2 R2-adj. R2-cv
## 1 net_opposition 32 28 0.967 0.963 NA
##
##
## Etas from analysis of variance
## Eta Eta_partial
## mean_estimate_fiscal 0.5917 0.955
## dk_fiscal 0.0698 0.357
## Muslim 0.0196 0.107MOD_summary(lm(net_opposition ~ mean_estimate_fiscal + dk_fiscal + Muslim, data = dk_fiscal_sub %>% no_SYR()))##
## ---- Model summary ----
## Model coefficients
## Beta SE CI_lower CI_upper
## mean_estimate_fiscal -1.0248 0.0686 -1.1656 -0.884
## dk_fiscal 0.0692 0.0742 -0.0830 0.221
## Muslim 0.0225 0.0551 -0.0906 0.136
##
##
## Model meta-data
## outcome N df R2 R2-adj. R2-cv
## 1 net_opposition 31 27 0.965 0.961 NA
##
##
## Etas from analysis of variance
## Eta Eta_partial
## mean_estimate_fiscal 0.5354 0.9445
## dk_fiscal 0.0334 0.1766
## Muslim 0.0146 0.0783#mediation models
#with SYR
mediation::mediate(
lm(mean_estimate_fiscal ~ dk_fiscal, data = dk_fiscal_sub),
lm(net_opposition ~ mean_estimate_fiscal + dk_fiscal, data = dk_fiscal_sub),
treat = "dk_fiscal", mediator = "mean_estimate_fiscal"
) %>% summary()##
## Causal Mediation Analysis
##
## Quasi-Bayesian Confidence Intervals
##
## Estimate 95% CI Lower 95% CI Upper p-value
## ACME -5.02e-03 -6.44e-03 0.00 <2e-16 ***
## ADE 6.55e-04 1.11e-05 0.00 0.048 *
## Total Effect -4.37e-03 -5.77e-03 0.00 <2e-16 ***
## Prop. Mediated 1.15e+00 1.00e+00 1.37 <2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Sample Size Used: 32
##
##
## Simulations: 1000#without SYR
mediation::mediate(
lm(mean_estimate_fiscal ~ dk_fiscal, data = dk_fiscal_sub %>% no_SYR()),
lm(net_opposition ~ mean_estimate_fiscal + dk_fiscal, data = dk_fiscal_sub %>% no_SYR()),
treat = "dk_fiscal", mediator = "mean_estimate_fiscal"
) %>% summary()##
## Causal Mediation Analysis
##
## Quasi-Bayesian Confidence Intervals
##
## Estimate 95% CI Lower 95% CI Upper p-value
## ACME -0.006434 -0.008083 0.00 <2e-16 ***
## ADE 0.000433 -0.000568 0.00 0.38
## Total Effect -0.006001 -0.007416 0.00 <2e-16 ***
## Prop. Mediated 1.071389 0.912137 1.27 <2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Sample Size Used: 31
##
##
## Simulations: 1000#cor matrix
cor_matrix(dk_fiscal_sub[c("dnk_net_fiscal_contribution_2014", "mean_estimate_fiscal", "Muslim", "net_opposition")],
CI = .95)## dnk_net_fiscal_contribution_2014
## dnk_net_fiscal_contribution_2014 "1"
## mean_estimate_fiscal "0.81 [0.64 0.90]"
## Muslim "-0.73 [-0.86 -0.51]"
## net_opposition "-0.75 [-0.87 -0.55]"
## mean_estimate_fiscal Muslim
## dnk_net_fiscal_contribution_2014 "0.81 [0.64 0.90]" "-0.73 [-0.86 -0.51]"
## mean_estimate_fiscal "1" "-0.72 [-0.86 -0.50]"
## Muslim "-0.72 [-0.86 -0.50]" "1"
## net_opposition "-0.98 [-0.99 -0.96]" "0.70 [0.47 0.85]"
## net_opposition
## dnk_net_fiscal_contribution_2014 "-0.75 [-0.87 -0.55]"
## mean_estimate_fiscal "-0.98 [-0.99 -0.96]"
## Muslim "0.70 [0.47 0.85]"
## net_opposition "1"cor_matrix(dk_fiscal_sub[c("dnk_net_fiscal_contribution_2014", "mean_estimate_fiscal", "Muslim", "net_opposition")] %>% no_SYR(),
CI = .95)## dnk_net_fiscal_contribution_2014
## dnk_net_fiscal_contribution_2014 "1"
## mean_estimate_fiscal "0.85 [0.70 0.92]"
## Muslim "-0.75 [-0.87 -0.54]"
## net_opposition "-0.82 [-0.91 -0.65]"
## mean_estimate_fiscal Muslim
## dnk_net_fiscal_contribution_2014 "0.85 [0.70 0.92]" "-0.75 [-0.87 -0.54]"
## mean_estimate_fiscal "1" "-0.70 [-0.84 -0.46]"
## Muslim "-0.70 [-0.84 -0.46]" "1"
## net_opposition "-0.98 [-0.99 -0.96]" "0.69 [0.44 0.84]"
## net_opposition
## dnk_net_fiscal_contribution_2014 "-0.82 [-0.91 -0.65]"
## mean_estimate_fiscal "-0.98 [-0.99 -0.96]"
## Muslim "0.69 [0.44 0.84]"
## net_opposition "1"Comparison with Carl 2016
Comparison with the UK findings.
#merge in UK data into main
dk_fiscal = full_join(dk_fiscal, noah16[c("ISO", "crime_rate_UK", "Log_incarceration_rate", "Log_crime_rate", "Net_opposition_UK")], by = "ISO")
#join DK prefs and accuracy into main
dk_fiscal = full_join(dk_fiscal, dk_fiscal_sub[c("ISO", "mean_estimate_fiscal", "net_opposition")], by = "ISO")
#correlations
dk_fiscal[c("dk_benefits_use", "dk_fiscal", "crime_rate_UK", "mean_estimate", "mean_estimate_fiscal", "net_opposition", "Net_opposition_UK", "Muslim")] %>%
set_colnames(c("DK: benefits use", "DK: net fiscal contribution", "UK: crime rate",
"DK: mean estimate benefits", "DK: mean estimate fiscal", "DK: Net opposition",
"UK: Net opposition", "Muslim% in home country")) %>%
cor_matrix()## DK: benefits use DK: net fiscal contribution
## DK: benefits use 1.000 -0.893
## DK: net fiscal contribution -0.893 1.000
## UK: crime rate 0.505 -0.410
## DK: mean estimate benefits 0.696 -0.891
## DK: mean estimate fiscal -0.724 0.806
## DK: Net opposition 0.676 -0.751
## UK: Net opposition 0.601 -0.849
## Muslim% in home country 0.680 -0.729
## UK: crime rate DK: mean estimate benefits
## DK: benefits use 0.505 0.696
## DK: net fiscal contribution -0.410 -0.891
## UK: crime rate 1.000 0.696
## DK: mean estimate benefits 0.696 1.000
## DK: mean estimate fiscal -0.699 -0.941
## DK: Net opposition 0.728 0.902
## UK: Net opposition 0.678 0.853
## Muslim% in home country 0.418 0.695
## DK: mean estimate fiscal DK: Net opposition
## DK: benefits use -0.724 0.676
## DK: net fiscal contribution 0.806 -0.751
## UK: crime rate -0.699 0.728
## DK: mean estimate benefits -0.941 0.902
## DK: mean estimate fiscal 1.000 -0.981
## DK: Net opposition -0.981 1.000
## UK: Net opposition -0.949 0.966
## Muslim% in home country -0.723 0.705
## UK: Net opposition Muslim% in home country
## DK: benefits use 0.601 0.680
## DK: net fiscal contribution -0.849 -0.729
## UK: crime rate 0.678 0.418
## DK: mean estimate benefits 0.853 0.695
## DK: mean estimate fiscal -0.949 -0.723
## DK: Net opposition 0.966 0.705
## UK: Net opposition 1.000 0.578
## Muslim% in home country 0.578 1.000dk_fiscal %>% filter(ISO != "SYR") %>% .[, c("dk_benefits_use", "dk_fiscal", "crime_rate_UK", "mean_estimate", "mean_estimate_fiscal", "net_opposition", "Net_opposition_UK", "Muslim")] %>%
set_colnames(c("DK: benefits use", "DK: net fiscal contribution", "UK: crime rate",
"DK: mean estimate benefits", "DK: mean estimate fiscal", "DK: Net opposition",
"UK: Net opposition", "Muslim% in home country")) %>%
cor_matrix()## DK: benefits use DK: net fiscal contribution
## DK: benefits use 1.000 -0.910
## DK: net fiscal contribution -0.910 1.000
## UK: crime rate 0.505 -0.410
## DK: mean estimate benefits 0.660 -0.887
## DK: mean estimate fiscal -0.695 0.847
## DK: Net opposition 0.658 -0.816
## UK: Net opposition 0.601 -0.849
## Muslim% in home country 0.660 -0.750
## UK: crime rate DK: mean estimate benefits
## DK: benefits use 0.505 0.660
## DK: net fiscal contribution -0.410 -0.887
## UK: crime rate 1.000 0.696
## DK: mean estimate benefits 0.696 1.000
## DK: mean estimate fiscal -0.699 -0.946
## DK: Net opposition 0.728 0.918
## UK: Net opposition 0.678 0.853
## Muslim% in home country 0.418 0.677
## DK: mean estimate fiscal DK: Net opposition
## DK: benefits use -0.695 0.658
## DK: net fiscal contribution 0.847 -0.816
## UK: crime rate -0.699 0.728
## DK: mean estimate benefits -0.946 0.918
## DK: mean estimate fiscal 1.000 -0.982
## DK: Net opposition -0.982 1.000
## UK: Net opposition -0.949 0.966
## Muslim% in home country -0.699 0.687
## UK: Net opposition Muslim% in home country
## DK: benefits use 0.601 0.660
## DK: net fiscal contribution -0.849 -0.750
## UK: crime rate 0.678 0.418
## DK: mean estimate benefits 0.853 0.677
## DK: mean estimate fiscal -0.949 -0.699
## DK: Net opposition 0.966 0.687
## UK: Net opposition 1.000 0.578
## Muslim% in home country 0.578 1.000#plot with UK net opposition
GG_scatter(dk_fiscal, "net_opposition", "Net_opposition_UK", case_names = "ISO") +
xlab("Net opposition: Denmark") +
ylab("Net opposition: UK") +
ggtitle("Net opposition to immigrants by country of origin",
"Danish data from study in review, UK data from YouGov")## `geom_smooth()` using formula 'y ~ x'
GG_save("figures/net_opposition_DK_UK.png")## `geom_smooth()` using formula 'y ~ x'Method variance
#compare the split half means
dk_fiscal_sub %>%
select(net_opposition, mean_estimate_fiscal, net_opposition_split1:mean_estimate_fiscal_split2) %>%
wtd.cors()## net_opposition mean_estimate_fiscal
## net_opposition 1.000 -0.981
## mean_estimate_fiscal -0.981 1.000
## net_opposition_split1 0.998 -0.977
## net_opposition_split2 0.998 -0.980
## mean_estimate_fiscal_split1 -0.978 0.999
## mean_estimate_fiscal_split2 -0.982 0.999
## net_opposition_split1 net_opposition_split2
## net_opposition 0.998 0.998
## mean_estimate_fiscal -0.977 -0.980
## net_opposition_split1 1.000 0.991
## net_opposition_split2 0.991 1.000
## mean_estimate_fiscal_split1 -0.975 -0.977
## mean_estimate_fiscal_split2 -0.977 -0.982
## mean_estimate_fiscal_split1
## net_opposition -0.978
## mean_estimate_fiscal 0.999
## net_opposition_split1 -0.975
## net_opposition_split2 -0.977
## mean_estimate_fiscal_split1 1.000
## mean_estimate_fiscal_split2 0.996
## mean_estimate_fiscal_split2
## net_opposition -0.982
## mean_estimate_fiscal 0.999
## net_opposition_split1 -0.977
## net_opposition_split2 -0.982
## mean_estimate_fiscal_split1 0.996
## mean_estimate_fiscal_split2 1.000Individual
Descriptives
#Our sample
#age
GG_denhist(d$age) +
xlab("Age")## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.
describe(d$age) %>% print()## vars n mean sd median trimmed mad min max range skew kurtosis se
## X1 1 476 39.3 13 40 39 16.3 18 65 47 0.102 -1.06 0.594#gender
table2(d$gender)#expected
folk1a_2017q2 %>%
filter(age %in% 18:65) %>%
{
c(
mean = wtd_mean(.$age, .$total),
SD = wtd_sd(.$age, .$total),
male = sum(.$men) / sum(.$total)
)
}## mean SD male
## 41.432 14.000 0.505#education distribution
table(d$education) %>% prop.table()##
## Grundskole
## 0.0906
## Gymnasial uddannelse (Alm. gymnasium/HF/HH/HGX/HTX)
## 0.1413
## Erhvervsuddannelser (HG, EFG, mesterlære m.v.)
## 0.2355
## Højere uddannelse af kort varighed (under 3 år)
## 0.1667
## Højere uddannelse af mellemlang varighed (3-4 år)
## 0.2246
## Højere uddannelse af lang varighed (5 år eller derover)
## 0.1413#expected is a little tricky, we have to merge but we can't exactly
HFUDD10_2017[-1] %>%
colSums() %>%
prop.table()## H10 Grundskole
## 0.20015
## H20 Gymnasiale uddannelser
## 0.11295
## H30 Erhvervsfaglige uddannelser
## 0.33007
## H40 Korte videregående uddannelser, KVU
## 0.05499
## H50 Mellemlange videregående uddannelser, MVU
## 0.15929
## H60 Bacheloruddannelser, BACH
## 0.02785
## H70 Lange videregående uddannelser, LVU
## 0.10513
## H80 Ph.d. og forskeruddannelser
## 0.00957#party voting intentions
table2(d$party_vote) %>% print(n = Inf)## # A tibble: 14 x 3
## Group Count Percent
## <chr> <dbl> <dbl>
## 1 Dansk Folkeparti 83 17.4
## 2 Socialdemokraterne 79 16.6
## 3 Venstre 53 11.1
## 4 Vote blank 45 9.45
## 5 Enhedslisten 36 7.56
## 6 Socialistisk Folkeparti 33 6.93
## 7 Liberal Alliance 29 6.09
## 8 Radikale Venstre 29 6.09
## 9 Alternativet 28 5.88
## 10 Would not vote 23 4.83
## 11 Nye Borgerlige 21 4.41
## 12 Konservative Folkeparti 11 2.31
## 13 Kristendemokraterne 6 1.26
## 14 <NA> 0 0#collection time
d$collected %>% mean()## [1] "2017-06-27"d$collected %>% range()## [1] "2017-05-28" "2017-07-19"d %>%
ggplot(aes(collected)) +
geom_histogram()## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.
#comparison
#https://en.wikipedia.org/wiki/Opinion_polling_for_the_next_Danish_general_election#2017
opinion_polls = tibble(
"Socialdemokraterne" = c(25.6, 25.3),
"Dansk Folkeparti" = c(17.4 , 20.9),
"Venstre" = c(20.3, 17.2),
"Enhedslisten" = c(7.4, 8.5),
"Liberal Alliance" = c(5.5, 6.1),
"Alternativet" = c(6.0, 5.7),
"Radikale Venstre" = c(5.6, 6.3),
"Socialistisk Folkeparti" = c(4.5, 5.2),
"Konservative Folkeparti" = c(4.5, 3.5),
"Kristendemokraterne" = c(0.6, 0.6),
"Nye Borgerlige" = c(2.2, 0.7)
) %>% t()
party_names = rownames(opinion_polls)
opinion_polls = opinion_polls %>% as_tibble() %>% mutate(avg_polls = rowMeans(.))## Warning: `as_tibble.matrix()` requires a matrix with column names or a `.name_repair` argument. Using compatibility `.name_repair`.
## This warning is displayed once per session.colnames(opinion_polls) = c("Gallup_July 6_2017", "Voxmeter_July_2_2017", "avg_polls")
opinion_polls$party_name = party_names
#add our values
our_poll = d %>% filter(!party_vote %in% c("Vil ikke stemme", "Stemme blankt")) %>% pull(party_vote) %>% table2()
colnames(our_poll) = c("party_name", "count", "this_sample")
left_join(opinion_polls, our_poll %>% select(party_name, this_sample)) %>%
GG_scatter("avg_polls", "this_sample", case_names = "party_name", repel_names = T)## Joining, by = "party_name"## `geom_smooth()` using formula 'y ~ x'
#economic estimates 2014 and 2015
GG_scatter(dk_fiscal, "dnk_net_fiscal_contribution_2014", "dnk_net_fiscal_contribution_2015", case_names = "ISO", repel_names = T) +
geom_abline(yintercept = 0, slope = 1, linetype = "dashed") +
xlab("Net fiscal contribution, 2014") +
ylab("Net fiscal contribution, 2015")## Warning: Ignoring unknown parameters: yintercept## `geom_smooth()` using formula 'y ~ x'
GG_save("figures/fiscal_contribs_2014_2015.png")## `geom_smooth()` using formula 'y ~ x'Accuracy
#sort dk_fiscal
dk_fiscal_sub = dplyr::arrange(dk_fiscal_sub, ISO)
#assert same order
assert_that(names(d_econ) %equals% names(d_prefs))## [1] TRUEassert_that(names(d_econ) %equals% dk_fiscal_sub$ISO)## [1] TRUE#score multiple metrics at a time
indi_accuracy = silence(score_accuracy(d_econ_num, criterion = dk_fiscal_sub$dnk_net_fiscal_contribution_2014, methods = c("pearson_r")))
indi_accuracy_noSYR = silence(score_accuracy(d_econ_num %>% no_SYR(), criterion = dk_fiscal_sub$dnk_net_fiscal_contribution_2014 %>% no_SYR(), methods = c("pearson_r")))
indi_accuracy %>% describe() %>% print()## vars n mean sd median trimmed mad min max range skew kurtosis
## X1 1 476 0.578 0.199 0.624 0.61 0.137 -0.451 0.874 1.33 -2.03 5.71
## se
## X1 0.00911indi_accuracy_noSYR %>% describe() %>% print()## vars n mean sd median trimmed mad min max range skew kurtosis
## X1 1 474 0.598 0.209 0.655 0.631 0.137 -0.538 0.878 1.42 -2.14 6.43
## se
## X1 0.0096#with syria
indi_accuracy %>%
GG_denhist(vline = "median") +
scale_x_continuous("Stereotype accuracy\nPearson r correlation with true values", breaks = seq(-1, 1, .1))## Warning in GG_denhist(., vline = "median"): received a data frame but no var:
## used the only available column## Scale for 'x' is already present. Adding another scale for 'x', which will
## replace the existing scale.## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.
GG_save("figures/indi_accu_dist.png")## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.#without
indi_accuracy_noSYR %>%
GG_denhist(vline = "median") +
scale_x_continuous("Stereotype accuracy\nPearson r correlation with true values", breaks = seq(-1, 1, .1)) +
labs(caption = "Syria excluded")## Warning in GG_denhist(., vline = "median"): received a data frame but no var:
## used the only available column## Warning: Removed 2 rows containing non-finite values (stat_bin).## Warning: Removed 2 rows containing non-finite values (stat_density).## Scale for 'x' is already present. Adding another scale for 'x', which will
## replace the existing scale.
## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.## Warning: Removed 2 rows containing non-finite values (stat_bin).
## Warning: Removed 2 rows containing non-finite values (stat_density).
GG_save("figures/indi_accu_dist_noSYR.png")## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.## Warning: Removed 2 rows containing non-finite values (stat_bin).
## Warning: Removed 2 rows containing non-finite values (stat_density).#cutoffs
wtd_mean(indi_accuracy$pearson_r >= .3)## [1] 0.912wtd_mean(indi_accuracy$pearson_r >= .5)## [1] 0.775wtd_mean(indi_accuracy_noSYR$pearson_r >= .3)## [1] 0.918wtd_mean(indi_accuracy_noSYR$pearson_r >= .5)## [1] 0.789#add metric to main data
d$stereotype_accuracy = indi_accuracy$pearson_r
d$stereotype_accuracy_noSYR = indi_accuracy_noSYR$pearson_rMuslim bias
#example case
example_muslim_bias = tibble(
Muslim = dk_fiscal_sub$Muslim,
true = dk_fiscal_sub$dk_fiscal,
estimate = d_econ_num[105, ] %>% unlist(),
name = dk_fiscal_sub$Names
)
#fit
example_muslim_bias_fit = lm(true ~ estimate, data = example_muslim_bias)
MOD_summary(example_muslim_bias_fit)##
## ---- Model summary ----
## Model coefficients
## Beta SE CI_lower CI_upper
## estimate 0.616 0.144 0.322 0.91
##
##
## Model meta-data
## outcome N df R2 R2-adj. R2-cv
## 1 true 32 30 0.379 0.359 NA
##
##
## Etas from analysis of variance
## Eta Eta_partial
## estimate 0.616 0.616example_muslim_bias$resid = resid(example_muslim_bias_fit)
#scatter
GG_scatter(example_muslim_bias, "estimate", "true", case_names = "name") +
scale_x_continuous("Estimated net economic contribution", breaks = 1:7) +
ylab("Net economic contribution")## `geom_smooth()` using formula 'y ~ x'
GG_scatter(example_muslim_bias, "Muslim", "resid", case_names = "name") +
scale_x_continuous("Muslim % in home country", labels = scales::percent) +
ylab("Residual of criterion ~ estimate")## `geom_smooth()` using formula 'y ~ x'
#adjust for Muslim bias
muslim_fit = lm(resid ~ Muslim, data = example_muslim_bias)
example_muslim_bias$estimate_adj = example_muslim_bias$estimate %>% standardize() + predict(muslim_fit) %>% standardize()
GG_scatter(example_muslim_bias, "estimate_adj", "true", case_names = "name") +
scale_x_continuous("Estimated net economic contribution") +
ylab("Net economic contribution")## `geom_smooth()` using formula 'y ~ x'
example_muslim_bias_fit2 = lm(true ~ estimate_adj, data = example_muslim_bias)
example_muslim_bias$resid2 = resid(example_muslim_bias_fit2)
GG_scatter(example_muslim_bias, "Muslim", "resid2", case_names = "name") +
scale_x_continuous("Muslim % in home country", labels = scales::percent) +
ylab("Residual of criterion ~ estimate")## `geom_smooth()` using formula 'y ~ x'
#calculate muslim bias per person
indi_Muslim_bias = plyr::adply(d_econ_num, .margins = 1, function(r) {
#local data
d_ = tibble(
true = dk_fiscal_sub$dk_fiscal,
estimate = r %>% unlist(),
Muslim = dk_fiscal_sub$Muslim
)
#fit accuracy
fit_accu = lm(true ~ estimate, data = d_)
#resids
d_$resids = resid(fit_accu) %>% standardize()
#out
tibble(
resid_x_Muslim = wtd.cors(d_$Muslim, d_$resids) %>% as.vector(),
resid_wtd_mean = wtd_mean(d_$resids, d_$Muslim)
)
})
indi_Muslim_bias_noSYR = plyr::adply(d_econ_num[-26], .margins = 1, function(r) {
#local data
d_ = tibble(
true = dk_fiscal_sub$dk_fiscal[-26],
estimate = r %>% unlist(),
Muslim = dk_fiscal_sub$Muslim[-26]
)
#fit accuracy
fit_accu = lm(true ~ estimate, data = d_)
#resids
d_$resids = resid(fit_accu) %>% standardize()
#out
tibble(
resid_x_Muslim = wtd.cors(d_$Muslim, d_$resids) %>% as.vector(),
resid_wtd_mean = wtd_mean(d_$resids, d_$Muslim)
)
})
#save to main
d$Muslim_bias_r = indi_Muslim_bias$resid_x_Muslim
d$Muslim_bias_wtd_mean = indi_Muslim_bias$resid_wtd_mean
d$Muslim_bias_r_noSYR = indi_Muslim_bias_noSYR$resid_x_Muslim
d$Muslim_bias_wtd_mean_noSYR = indi_Muslim_bias_noSYR$resid_wtd_mean
#compare metrics
GG_scatter(d, "Muslim_bias_r", "Muslim_bias_wtd_mean") +
xlab("Muslim bias (correlation metric)") +
ylab("Muslim bias (weighted mean residual metric)")## `geom_smooth()` using formula 'y ~ x'
#distribution
describe(d$Muslim_bias_r)## vars n mean sd median trimmed mad min max range skew
## X1 1 476 -0.486 0.161 -0.494 -0.494 0.166 -0.756 0.0307 0.787 0.463
## kurtosis se
## X1 -0.159 0.00737sum(d$Muslim_bias_r > 0)## [1] 1describe(d$Muslim_bias_r_noSYR)## vars n mean sd median trimmed mad min max range skew kurtosis
## X1 1 476 -0.528 0.162 -0.547 -0.54 0.164 -0.805 0.014 0.819 0.67 0.302
## se
## X1 0.00743sum(d$Muslim_bias_r_noSYR > 0)## [1] 2GG_denhist(d, "Muslim_bias_r") +
xlab("Muslim bias (correlation metric)")## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.
GG_save("figures/muslim_bias_dist.png")## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.GG_denhist(d, "Muslim_bias_r_noSYR") +
xlab("Muslim bias (correlation metric)") +
labs(caption = "Syria excluded")## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.
GG_save("figures/muslim_bias_dist_noSYR.png")## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.Mediation analysis
#example
mediation_example = tibble(
preference = d_prefs_num_alt[146, ] %>% unlist(),
stereotype = d_econ_num[146, ] %>% unlist(),
Muslim = dk_fiscal_sub$Muslim,
name = dk_fiscal_sub$Names
)
#fit
(mediation_example_ORM = rms::orm(preference ~ stereotype, data = mediation_example))## Logistic (Proportional Odds) Ordinal Regression Model
##
## rms::orm(formula = preference ~ stereotype, data = mediation_example)
##
##
## Frequencies of Responses
##
## 1 2 3 4
## 5 10 16 1
##
## Model Likelihood Discrimination Rank Discrim.
## Ratio Test Indexes Indexes
## Obs 32 LR chi2 19.48 R2 0.512 rho 0.666
## Distinct Y 4 d.f. 1 g 2.207
## Median Y 3 Pr(> chi2) <0.0001 gr 9.088
## max |deriv| 2e-05 Score chi2 18.34 |Pr(Y>=median)-0.5| 0.312
## Pr(> chi2) <0.0001
##
## Coef S.E. Wald Z Pr(>|Z|)
## y>=2 -2.0647 1.0707 -1.93 0.0538
## y>=3 -4.6738 1.3593 -3.44 0.0006
## y>=4 -9.4871 2.0820 -4.56 <0.0001
## stereotype 1.0753 0.2924 3.68 0.0002
## (mediation_example_OLS = rms::ols(preference ~ stereotype, data = mediation_example))## Linear Regression Model
##
## rms::ols(formula = preference ~ stereotype, data = mediation_example)
##
## Model Likelihood Discrimination
## Ratio Test Indexes
## Obs 32 LR chi2 18.12 R2 0.432
## sigma0.6108 d.f. 1 R2 adj 0.413
## d.f. 30 Pr(> chi2) 0.0000 g 0.601
##
## Residuals
##
## Min 1Q Median 3Q Max
## -1.296482 -0.296482 -0.003769 0.118090 1.996231
##
##
## Coef S.E. t Pr(>|t|)
## Intercept 1.1256 0.2889 3.90 0.0005
## stereotype 0.2927 0.0612 4.78 <0.0001
## #plots
mediation_example$resid = resid(mediation_example_OLS)
GG_scatter(mediation_example, "stereotype", "preference", case_names = "name", repel_names = T) +
scale_x_continuous(breaks = 1:7) +
scale_y_continuous(breaks = 1:4, limit = c(NA, 4.5))## `geom_smooth()` using formula 'y ~ x'
GG_scatter(mediation_example, "Muslim", "resid", case_names = "name", repel_names = T) +
scale_x_continuous("Muslim % in home country", labels = scales::percent) +
ylab("Residual of preference ~ stereotype")## `geom_smooth()` using formula 'y ~ x'
#stereotypes and preferences
mediation_model = plyr::ldply(seq_along_rows(d), function(ri) {
#collect data
d_ = tibble(
preference = d_prefs_num_alt[ri, ] %>% unlist(),
stereotype = d_econ_num[ri, ] %>% unlist(),
Muslim = dk_fiscal_sub$Muslim,
preference_ordinal = ordered(preference),
stereotype_ordinal = ordered(stereotype)
)
#out
out = tibble(
# mediation_ORM = NA,
mediation_r = NA,
mediation_lr = NA,
# Muslim_resid_ORM = NA,
Muslim_resid_OLS = NA
)
#fit models
# ORM = try({
# ORM_fit = silence(rms::orm(preference ~ stereotype, data = d_))
# out$Muslim_resid_ORM = silence(wtd.cors(resid(ORM_fit), d_$Muslim) %>% as.vector())
# out$mediation_ORM = ORM_fit$stats[11] %>% sqrt()
# }, silent = T)
OLS = try({
OLS_fit = silence(rms::ols(preference ~ stereotype, data = d_))
out$Muslim_resid_OLS = silence(wtd.cors(resid(OLS_fit), d_$Muslim) %>% as.vector() %>% Inf_to_NA())
out$mediation_r = wtd.cors(d_$preference, d_$stereotype) %>% as.vector() %>% Inf_to_NA()
}, silent = T)
silence({LC = try({
#latent correlations
lc = hetcor(as.data.frame(d_[c(4, 5, 3)])) %>% .$correlations
#mediation latent
out$mediation_lr = lc %>% .[2, 1]
#partial correlation
out$Muslim_partial = partial.r(lc, c(1, 3), 2) %>% .[2, 1]
}, silent = T)})
out
})
rownames(mediation_model) = 1:nrow(mediation_model)
#no SYR
mediation_model_noSYR = plyr::ldply(seq_along_rows(d), function(ri) {
#collect data
d_ = tibble(
preference = d_prefs_num_alt[ri, -26] %>% unlist(),
stereotype = d_econ_num[ri, -26] %>% unlist(),
Muslim = dk_fiscal_sub$Muslim[-26],
preference_ordinal = ordered(preference),
stereotype_ordinal = ordered(stereotype)
)
#out
out = tibble(
# mediation_ORM = NA,
mediation_r = NA,
mediation_lr = NA,
# Muslim_resid_ORM = NA,
Muslim_resid_OLS = NA
)
#fit models
# ORM = try({
# ORM_fit = silence(rms::orm(preference ~ stereotype, data = d_))
# out$Muslim_resid_ORM = silence(wtd.cors(resid(ORM_fit), d_$Muslim) %>% as.vector())
# out$mediation_ORM = ORM_fit$stats[11] %>% sqrt()
# }, silent = T)
OLS = try({
OLS_fit = silence(rms::ols(preference ~ stereotype, data = d_))
out$Muslim_resid_OLS = silence(wtd.cors(resid(OLS_fit), d_$Muslim) %>% as.vector() %>% Inf_to_NA())
out$mediation_r = wtd.cors(d_$preference, d_$stereotype) %>% as.vector() %>% Inf_to_NA()
}, silent = T)
silence({LC = try({
#latent correlations
lc = hetcor(as.data.frame(d_[c(4, 5, 3)])) %>% .$correlations
#mediation latent
out$mediation_lr = lc %>% .[2, 1]
#partial correlation
out$Muslim_partial = partial.r(lc, c(1, 3), 2) %>% .[2, 1]
}, silent = T)})
out
})
rownames(mediation_model_noSYR) = 1:nrow(mediation_model_noSYR)
#stats
describe(mediation_model$mediation_r) %>% print()## vars n mean sd median trimmed mad min max range skew kurtosis
## X1 1 438 0.567 0.342 0.668 0.625 0.229 -0.804 0.979 1.78 -1.71 2.94
## se
## X1 0.0163describe(mediation_model$mediation_lr) %>% print()## vars n mean sd median trimmed mad min max range skew kurtosis
## X1 1 380 0.634 0.412 0.777 0.724 0.204 -0.916 1 1.92 -2.07 3.92
## se
## X1 0.0212describe(mediation_model$Muslim_resid_OLS) %>% print()## vars n mean sd median trimmed mad min max range skew kurtosis
## X1 1 476 -0.109 0.246 -0.129 -0.114 0.241 -0.788 0.632 1.42 0.212 -0.234
## se
## X1 0.0113describe(mediation_model$Muslim_partial) %>% print()## vars n mean sd median trimmed mad min max range skew kurtosis
## X1 1 368 -0.21 0.423 -0.272 -0.233 0.422 -1 1 2 0.533 0.0408
## se
## X1 0.0221describe(mediation_model_noSYR$mediation_r) %>% print()## vars n mean sd median trimmed mad min max range skew kurtosis
## X1 1 435 0.569 0.335 0.661 0.627 0.227 -0.801 0.978 1.78 -1.75 3.18
## se
## X1 0.0161describe(mediation_model_noSYR$mediation_lr) %>% print()## vars n mean sd median trimmed mad min max range skew kurtosis
## X1 1 376 0.645 0.399 0.78 0.731 0.195 -0.913 1 1.91 -2.12 4.26
## se
## X1 0.0206describe(mediation_model_noSYR$Muslim_resid_OLS) %>% print()## vars n mean sd median trimmed mad min max range skew kurtosis
## X1 1 449 -0.135 0.235 -0.15 -0.142 0.222 -0.786 0.592 1.38 0.257 -0.00304
## se
## X1 0.0111describe(mediation_model_noSYR$Muslim_partial) %>% print()## vars n mean sd median trimmed mad min max range skew kurtosis
## X1 1 362 -0.218 0.424 -0.266 -0.242 0.41 -1 1 2 0.577 0.188
## se
## X1 0.0223#cors
wtd.cors(mediation_model)## mediation_r mediation_lr Muslim_resid_OLS Muslim_partial
## mediation_r 1.000 0.956 -0.282 -0.342
## mediation_lr 0.956 1.000 -0.340 -0.344
## Muslim_resid_OLS -0.282 -0.340 1.000 0.827
## Muslim_partial -0.342 -0.344 0.827 1.000wtd.cors(mediation_model_noSYR)## mediation_r mediation_lr Muslim_resid_OLS Muslim_partial
## mediation_r 1.000 0.955 -0.239 -0.291
## mediation_lr 0.955 1.000 -0.291 -0.282
## Muslim_resid_OLS -0.239 -0.291 1.000 0.818
## Muslim_partial -0.291 -0.282 0.818 1.000#dists
GG_denhist(mediation_model, "mediation_r", vline = median) +
xlab("Strength of mediation (correlation)")## Warning: Removed 38 rows containing non-finite values (stat_bin).## Warning: Removed 38 rows containing non-finite values (stat_density).## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.## Warning: Removed 38 rows containing non-finite values (stat_bin).
## Warning: Removed 38 rows containing non-finite values (stat_density).
GG_denhist(mediation_model, "Muslim_resid_OLS", vline = median) +
xlab("Correlation of Muslim % with residual from preference ~ stereotype")## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.
GG_denhist(mediation_model, "Muslim_partial", vline = median) +
xlab("Partial correlation of Muslim % with preference holding stereotype constant")## Warning: Removed 108 rows containing non-finite values (stat_bin).## Warning: Removed 108 rows containing non-finite values (stat_density).## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.## Warning: Removed 108 rows containing non-finite values (stat_bin).
## Warning: Removed 108 rows containing non-finite values (stat_density).
#no SYR
GG_denhist(mediation_model_noSYR, "mediation_r", vline = median) +
xlab("Strength of mediation (correlation)")## Warning: Removed 41 rows containing non-finite values (stat_bin).## Warning: Removed 41 rows containing non-finite values (stat_density).## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.## Warning: Removed 41 rows containing non-finite values (stat_bin).
## Warning: Removed 41 rows containing non-finite values (stat_density).
GG_denhist(mediation_model_noSYR, "Muslim_resid_OLS", vline = median) +
xlab("Correlation of Muslim % with residual from preference ~ stereotype")## Warning: Removed 27 rows containing non-finite values (stat_bin).## Warning: Removed 27 rows containing non-finite values (stat_density).## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.## Warning: Removed 27 rows containing non-finite values (stat_bin).
## Warning: Removed 27 rows containing non-finite values (stat_density).
GG_denhist(mediation_model_noSYR, "Muslim_partial", vline = median) +
xlab("Partial correlation of Muslim % with preference holding stereotype constant")## Warning: Removed 114 rows containing non-finite values (stat_bin).## Warning: Removed 114 rows containing non-finite values (stat_density).## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.## Warning: Removed 114 rows containing non-finite values (stat_bin).
## Warning: Removed 114 rows containing non-finite values (stat_density).
#add to main data
d$Muslim_preference = mediation_model$Muslim_partial
d$Muslim_preference_noSYR = mediation_model_noSYR$Muslim_partial
#describe
d$Muslim_preference %>% describe()## vars n mean sd median trimmed mad min max range skew kurtosis
## X1 1 368 -0.21 0.423 -0.272 -0.233 0.422 -1 1 2 0.533 0.0408
## se
## X1 0.0221d$Muslim_preference_noSYR %>% describe()## vars n mean sd median trimmed mad min max range skew kurtosis
## X1 1 362 -0.218 0.424 -0.266 -0.242 0.41 -1 1 2 0.577 0.188
## se
## X1 0.0223Predictors of individual primary outcomes
#define var groups
main_outcomes = c("stereotype_accuracy", "Muslim_bias_r", "Muslim_preference")
main_outcomes_noSYR = c("stereotype_accuracy_noSYR", "Muslim_bias_r_noSYR", "Muslim_preference_noSYR")
predictors = c("age", "education_num", "Muslims_are_treated_well", "DK_admit_only_net_positive_immigrants", "nonwesterns_are_net_positive")
#cors
wtd.cors(d[c(main_outcomes, predictors)])## stereotype_accuracy Muslim_bias_r
## stereotype_accuracy 1.0000 0.7424
## Muslim_bias_r 0.7424 1.0000
## Muslim_preference -0.0437 -0.0516
## age 0.1397 0.0897
## education_num 0.1778 0.1929
## Muslims_are_treated_well 0.0855 0.1021
## DK_admit_only_net_positive_immigrants 0.1536 0.1791
## nonwesterns_are_net_positive -0.1244 -0.1593
## Muslim_preference age education_num
## stereotype_accuracy -0.0437 0.1397 0.1778
## Muslim_bias_r -0.0516 0.0897 0.1929
## Muslim_preference 1.0000 -0.1179 0.0173
## age -0.1179 1.0000 -0.0519
## education_num 0.0173 -0.0519 1.0000
## Muslims_are_treated_well -0.1864 0.1283 -0.0917
## DK_admit_only_net_positive_immigrants -0.3146 0.0981 -0.1099
## nonwesterns_are_net_positive 0.0888 -0.1106 0.1505
## Muslims_are_treated_well
## stereotype_accuracy 0.0855
## Muslim_bias_r 0.1021
## Muslim_preference -0.1864
## age 0.1283
## education_num -0.0917
## Muslims_are_treated_well 1.0000
## DK_admit_only_net_positive_immigrants 0.4928
## nonwesterns_are_net_positive -0.1930
## DK_admit_only_net_positive_immigrants
## stereotype_accuracy 0.1536
## Muslim_bias_r 0.1791
## Muslim_preference -0.3146
## age 0.0981
## education_num -0.1099
## Muslims_are_treated_well 0.4928
## DK_admit_only_net_positive_immigrants 1.0000
## nonwesterns_are_net_positive -0.2130
## nonwesterns_are_net_positive
## stereotype_accuracy -0.1244
## Muslim_bias_r -0.1593
## Muslim_preference 0.0888
## age -0.1106
## education_num 0.1505
## Muslims_are_treated_well -0.1930
## DK_admit_only_net_positive_immigrants -0.2130
## nonwesterns_are_net_positive 1.0000wtd.cors(d[c(main_outcomes_noSYR, predictors)])## stereotype_accuracy_noSYR
## stereotype_accuracy_noSYR 1.00000
## Muslim_bias_r_noSYR 0.68823
## Muslim_preference_noSYR -0.00366
## age 0.17603
## education_num 0.16281
## Muslims_are_treated_well 0.08657
## DK_admit_only_net_positive_immigrants 0.14437
## nonwesterns_are_net_positive -0.12769
## Muslim_bias_r_noSYR
## stereotype_accuracy_noSYR 0.6882
## Muslim_bias_r_noSYR 1.0000
## Muslim_preference_noSYR -0.0254
## age 0.1059
## education_num 0.1878
## Muslims_are_treated_well 0.1040
## DK_admit_only_net_positive_immigrants 0.1807
## nonwesterns_are_net_positive -0.1612
## Muslim_preference_noSYR age
## stereotype_accuracy_noSYR -0.00366 0.1760
## Muslim_bias_r_noSYR -0.02540 0.1059
## Muslim_preference_noSYR 1.00000 -0.1161
## age -0.11608 1.0000
## education_num 0.03314 -0.0519
## Muslims_are_treated_well -0.17431 0.1283
## DK_admit_only_net_positive_immigrants -0.29203 0.0981
## nonwesterns_are_net_positive 0.08461 -0.1106
## education_num Muslims_are_treated_well
## stereotype_accuracy_noSYR 0.1628 0.0866
## Muslim_bias_r_noSYR 0.1878 0.1040
## Muslim_preference_noSYR 0.0331 -0.1743
## age -0.0519 0.1283
## education_num 1.0000 -0.0917
## Muslims_are_treated_well -0.0917 1.0000
## DK_admit_only_net_positive_immigrants -0.1099 0.4928
## nonwesterns_are_net_positive 0.1505 -0.1930
## DK_admit_only_net_positive_immigrants
## stereotype_accuracy_noSYR 0.1444
## Muslim_bias_r_noSYR 0.1807
## Muslim_preference_noSYR -0.2920
## age 0.0981
## education_num -0.1099
## Muslims_are_treated_well 0.4928
## DK_admit_only_net_positive_immigrants 1.0000
## nonwesterns_are_net_positive -0.2130
## nonwesterns_are_net_positive
## stereotype_accuracy_noSYR -0.1277
## Muslim_bias_r_noSYR -0.1612
## Muslim_preference_noSYR 0.0846
## age -0.1106
## education_num 0.1505
## Muslims_are_treated_well -0.1930
## DK_admit_only_net_positive_immigrants -0.2130
## nonwesterns_are_net_positive 1.0000#plot
GG_scatter(d, "stereotype_accuracy", "Muslim_bias_r") +
geom_smooth(se = F) +
scale_x_continuous("Stereotype accuracy\nCorrelation of estimates with criterion data", breaks = seq(-1, 1, .2)) +
ylab("Muslim bias in estimate\nCorrelation of estimate residual with Muslim%")## `geom_smooth()` using formula 'y ~ x'## `geom_smooth()` using method = 'loess' and formula 'y ~ x'
GG_save("figures/accuracy_bias.png")## `geom_smooth()` using formula 'y ~ x'
## `geom_smooth()` using method = 'loess' and formula 'y ~ x'GG_scatter(d, "stereotype_accuracy_noSYR", "Muslim_bias_r_noSYR") +
geom_smooth(se = F) +
scale_x_continuous("Stereotype accuracy\nCorrelation of estimates with criterion data", breaks = seq(-1, 1, .2)) +
ylab("Muslim bias in estimate\nCorrelation of estimate residual with Muslim%") +
labs(caption = "Syria excluded")## `geom_smooth()` using formula 'y ~ x'
## `geom_smooth()` using method = 'loess' and formula 'y ~ x'
GG_save("figures/accuracy_bias_noSYR.png")## `geom_smooth()` using formula 'y ~ x'
## `geom_smooth()` using method = 'loess' and formula 'y ~ x'#models
#with Syria
list(
rms::ols(stereotype_accuracy ~ age + gender, data = d %>% df_standardize(messages = F)),
rms::ols(stereotype_accuracy ~ age + gender + education_num, data = d %>% df_standardize(messages = F)),
rms::ols(stereotype_accuracy ~ age + gender + block_vote, data = d %>% df_standardize(messages = F)),
rms::ols(stereotype_accuracy ~ age + gender + education_num + block_vote, data = d %>% df_standardize(messages = F)),
rms::ols(stereotype_accuracy ~ age + gender + party_vote, data = d %>% df_standardize(messages = F)),
rms::ols(stereotype_accuracy ~ age + gender + education_num + party_vote, data = d %>% df_standardize(messages = F))
) %>%
summarize_models(asterisks = c(.05, .01, .005)) %>%
DT::datatable()#without
list(
rms::ols(stereotype_accuracy_noSYR ~ age + gender, data = d %>% df_standardize(messages = F)),
rms::ols(stereotype_accuracy_noSYR ~ age + gender + education_num, data = d %>% df_standardize(messages = F)),
rms::ols(stereotype_accuracy_noSYR ~ age + gender + block_vote, data = d %>% df_standardize(messages = F)),
rms::ols(stereotype_accuracy_noSYR ~ age + gender + education_num + block_vote, data = d %>% df_standardize(messages = F)),
rms::ols(stereotype_accuracy_noSYR ~ age + gender + party_vote, data = d %>% df_standardize(messages = F)),
rms::ols(stereotype_accuracy_noSYR ~ age + gender + education_num + party_vote, data = d %>% df_standardize(messages = F))
) %>%
summarize_models(asterisks = c(.05, .01, .005)) %>%
DT::datatable()#not standardized
#models
#with Syria
list(
rms::ols(stereotype_accuracy ~ age + gender, data = d),
rms::ols(stereotype_accuracy ~ age + gender + education_num, data = d),
rms::ols(stereotype_accuracy ~ age + gender + block_vote, data = d),
rms::ols(stereotype_accuracy ~ age + gender + education_num + block_vote, data = d),
rms::ols(stereotype_accuracy ~ age + gender + party_vote, data = d),
rms::ols(stereotype_accuracy ~ age + gender + education_num + party_vote, data = d)
) %>%
summarize_models(beta_digits = 4, asterisks = c(.05, .01, .005)) %>%
DT::datatable()#without
list(
rms::ols(stereotype_accuracy_noSYR ~ age + gender, data = d),
rms::ols(stereotype_accuracy_noSYR ~ age + gender + education_num, data = d),
rms::ols(stereotype_accuracy_noSYR ~ age + gender + block_vote, data = d),
rms::ols(stereotype_accuracy_noSYR ~ age + gender + education_num + block_vote, data = d),
rms::ols(stereotype_accuracy_noSYR ~ age + gender + party_vote, data = d),
rms::ols(stereotype_accuracy_noSYR ~ age + gender + education_num + party_vote, data = d)
) %>%
summarize_models(beta_digits = 4, asterisks = c(.05, .01, .005)) %>%
DT::datatable()#party models
#with SYR
rms::ols(stereotype_accuracy ~ party_vote, data = d)## Linear Regression Model
##
## rms::ols(formula = stereotype_accuracy ~ party_vote, data = d)
##
## Model Likelihood Discrimination
## Ratio Test Indexes
## Obs 476 LR chi2 16.57 R2 0.034
## sigma0.1977 d.f. 12 R2 adj 0.009
## d.f. 463 Pr(> chi2) 0.1665 g 0.041
##
## Residuals
##
## Min 1Q Median 3Q Max
## -1.02772 -0.06558 0.04243 0.13307 0.31234
##
##
## Coef S.E. t Pr(>|t|)
## Intercept 0.5884 0.0222 26.45 <0.0001
## party_vote=Alternativet -0.0471 0.0435 -1.08 0.2793
## party_vote=Dansk Folkeparti -0.0115 0.0311 -0.37 0.7127
## party_vote=Enhedslisten -0.0284 0.0398 -0.71 0.4761
## party_vote=Konservative Folkeparti 0.0705 0.0636 1.11 0.2687
## party_vote=Kristendemokraterne 0.0776 0.0837 0.93 0.3544
## party_vote=Liberal Alliance 0.0613 0.0429 1.43 0.1543
## party_vote=Nye Borgerlige -0.0154 0.0485 -0.32 0.7510
## party_vote=Radikale Venstre -0.0263 0.0429 -0.61 0.5408
## party_vote=Socialistisk Folkeparti -0.0689 0.0410 -1.68 0.0935
## party_vote=Venstre 0.0357 0.0351 1.02 0.3094
## party_vote=Vote blank -0.0477 0.0369 -1.29 0.1973
## party_vote=Would not vote -0.0399 0.0469 -0.85 0.3952
## rms::ols(Muslim_bias_r ~ party_vote, data = d)## Linear Regression Model
##
## rms::ols(formula = Muslim_bias_r ~ party_vote, data = d)
##
## Model Likelihood Discrimination
## Ratio Test Indexes
## Obs 476 LR chi2 24.43 R2 0.050
## sigma0.1586 d.f. 12 R2 adj 0.025
## d.f. 463 Pr(> chi2) 0.0178 g 0.041
##
## Residuals
##
## Min 1Q Median 3Q Max
## -0.30640 -0.12056 -0.01092 0.10114 0.47576
##
##
## Coef S.E. t Pr(>|t|)
## Intercept -0.5024 0.0178 -28.15 <0.0001
## party_vote=Alternativet -0.0324 0.0349 -0.93 0.3541
## party_vote=Dansk Folkeparti 0.0452 0.0249 1.81 0.0708
## party_vote=Enhedslisten -0.0272 0.0319 -0.85 0.3944
## party_vote=Konservative Folkeparti 0.0813 0.0511 1.59 0.1121
## party_vote=Kristendemokraterne 0.0805 0.0672 1.20 0.2314
## party_vote=Liberal Alliance 0.0855 0.0344 2.48 0.0134
## party_vote=Nye Borgerlige 0.0499 0.0389 1.28 0.2006
## party_vote=Radikale Venstre 0.0140 0.0344 0.41 0.6838
## party_vote=Socialistisk Folkeparti -0.0212 0.0329 -0.64 0.5204
## party_vote=Venstre 0.0394 0.0282 1.40 0.1629
## party_vote=Vote blank -0.0260 0.0296 -0.88 0.3808
## party_vote=Would not vote 0.0185 0.0376 0.49 0.6221
## rms::ols(Muslim_preference ~ party_vote, data = d)## Frequencies of Missing Values Due to Each Variable
## Muslim_preference party_vote
## 108 0
##
## Linear Regression Model
##
## rms::ols(formula = Muslim_preference ~ party_vote, data = d)
##
##
## Model Likelihood Discrimination
## Ratio Test Indexes
## Obs 368 LR chi2 24.64 R2 0.065
## sigma0.4164 d.f. 12 R2 adj 0.033
## d.f. 355 Pr(> chi2) 0.0166 g 0.097
##
## Residuals
##
## Min 1Q Median 3Q Max
## -0.93362 -0.27710 -0.03843 0.27817 1.28037
##
##
## Coef S.E. t Pr(>|t|)
## Intercept -0.2526 0.0533 -4.74 <0.0001
## party_vote=Alternativet 0.3989 0.1073 3.72 0.0002
## party_vote=Dansk Folkeparti -0.0278 0.0742 -0.37 0.7084
## party_vote=Enhedslisten 0.0627 0.0950 0.66 0.5100
## party_vote=Konservative Folkeparti 0.0857 0.1420 0.60 0.5466
## party_vote=Kristendemokraterne -0.1242 0.2149 -0.58 0.5637
## party_vote=Liberal Alliance 0.0888 0.1035 0.86 0.3914
## party_vote=Nye Borgerlige 0.0250 0.1142 0.22 0.8267
## party_vote=Radikale Venstre 0.2321 0.1035 2.24 0.0256
## party_vote=Socialistisk Folkeparti 0.0133 0.1003 0.13 0.8949
## party_vote=Venstre -0.0287 0.0808 -0.36 0.7225
## party_vote=Vote blank 0.0306 0.0909 0.34 0.7369
## party_vote=Would not vote -0.0088 0.1169 -0.08 0.9399
## #without
rms::ols(stereotype_accuracy_noSYR ~ party_vote, data = d)## Frequencies of Missing Values Due to Each Variable
## stereotype_accuracy_noSYR party_vote
## 2 0
##
## Linear Regression Model
##
## rms::ols(formula = stereotype_accuracy_noSYR ~ party_vote, data = d)
##
##
## Model Likelihood Discrimination
## Ratio Test Indexes
## Obs 474 LR chi2 19.30 R2 0.040
## sigma0.2074 d.f. 12 R2 adj 0.015
## d.f. 461 Pr(> chi2) 0.0815 g 0.047
##
## Residuals
##
## Min 1Q Median 3Q Max
## -1.14052 -0.06976 0.05080 0.13141 0.31223
##
##
## Coef S.E. t Pr(>|t|)
## Intercept 0.6160 0.0233 26.40 <0.0001
## party_vote=Alternativet -0.0574 0.0462 -1.24 0.2153
## party_vote=Dansk Folkeparti -0.0130 0.0326 -0.40 0.6891
## party_vote=Enhedslisten -0.0266 0.0417 -0.64 0.5246
## party_vote=Konservative Folkeparti 0.0366 0.0667 0.55 0.5838
## party_vote=Kristendemokraterne 0.0896 0.0878 1.02 0.3081
## party_vote=Liberal Alliance 0.0545 0.0450 1.21 0.2267
## party_vote=Nye Borgerlige -0.0585 0.0519 -1.13 0.2600
## party_vote=Radikale Venstre -0.0318 0.0450 -0.71 0.4806
## party_vote=Socialistisk Folkeparti -0.0909 0.0430 -2.12 0.0350
## party_vote=Venstre 0.0362 0.0368 0.98 0.3257
## party_vote=Vote blank -0.0621 0.0387 -1.60 0.1096
## party_vote=Would not vote -0.0597 0.0491 -1.22 0.2250
## rms::ols(Muslim_bias_r_noSYR ~ party_vote, data = d)## Linear Regression Model
##
## rms::ols(formula = Muslim_bias_r_noSYR ~ party_vote, data = d)
##
## Model Likelihood Discrimination
## Ratio Test Indexes
## Obs 476 LR chi2 26.77 R2 0.055
## sigma0.1596 d.f. 12 R2 adj 0.030
## d.f. 463 Pr(> chi2) 0.0083 g 0.043
##
## Residuals
##
## Min 1Q Median 3Q Max
## -0.30145 -0.12258 -0.01139 0.09631 0.52648
##
##
## Coef S.E. t Pr(>|t|)
## Intercept -0.5428 0.0180 -30.22 <0.0001
## party_vote=Alternativet -0.0435 0.0351 -1.24 0.2158
## party_vote=Dansk Folkeparti 0.0475 0.0251 1.90 0.0587
## party_vote=Enhedslisten -0.0308 0.0321 -0.96 0.3378
## party_vote=Konservative Folkeparti 0.0721 0.0514 1.40 0.1611
## party_vote=Kristendemokraterne 0.0735 0.0676 1.09 0.2771
## party_vote=Liberal Alliance 0.0847 0.0347 2.45 0.0148
## party_vote=Nye Borgerlige 0.0576 0.0392 1.47 0.1422
## party_vote=Radikale Venstre 0.0166 0.0347 0.48 0.6317
## party_vote=Socialistisk Folkeparti -0.0296 0.0331 -0.90 0.3709
## party_vote=Venstre 0.0344 0.0283 1.21 0.2253
## party_vote=Vote blank -0.0280 0.0298 -0.94 0.3474
## party_vote=Would not vote 0.0200 0.0378 0.53 0.5964
## rms::ols(Muslim_preference_noSYR ~ party_vote, data = d)## Frequencies of Missing Values Due to Each Variable
## Muslim_preference_noSYR party_vote
## 114 0
##
## Linear Regression Model
##
## rms::ols(formula = Muslim_preference_noSYR ~ party_vote, data = d)
##
##
## Model Likelihood Discrimination
## Ratio Test Indexes
## Obs 362 LR chi2 22.55 R2 0.060
## sigma0.4184 d.f. 12 R2 adj 0.028
## d.f. 349 Pr(> chi2) 0.0318 g 0.094
##
## Residuals
##
## Min 1Q Median 3Q Max
## -0.98054 -0.27540 -0.02135 0.27211 1.28241
##
##
## Coef S.E. t Pr(>|t|)
## Intercept -0.2824 0.0545 -5.19 <0.0001
## party_vote=Alternativet 0.4084 0.1127 3.62 0.0003
## party_vote=Dansk Folkeparti 0.0027 0.0752 0.04 0.9715
## party_vote=Enhedslisten 0.0861 0.0960 0.90 0.3702
## party_vote=Konservative Folkeparti 0.1140 0.1497 0.76 0.4468
## party_vote=Kristendemokraterne -0.0872 0.2162 -0.40 0.6871
## party_vote=Liberal Alliance 0.1227 0.1045 1.17 0.2414
## party_vote=Nye Borgerlige 0.0496 0.1152 0.43 0.6668
## party_vote=Radikale Venstre 0.2630 0.1045 2.52 0.0123
## party_vote=Socialistisk Folkeparti 0.0240 0.1028 0.23 0.8160
## party_vote=Venstre 0.0037 0.0818 0.05 0.9640
## party_vote=Vote blank 0.0703 0.0918 0.77 0.4446
## party_vote=Would not vote 0.0291 0.1179 0.25 0.8052
## #cohen d
SMD_matrix(d$stereotype_accuracy, d$party_vote)## Socialdemokraterne Alternativet Dansk Folkeparti
## Socialdemokraterne NA 0.23824 0.0579
## Alternativet 0.2382 NA -0.1803
## Dansk Folkeparti 0.0579 -0.18033 NA
## Enhedslisten 0.1434 -0.09484 0.0855
## Konservative Folkeparti -0.3564 -0.59463 -0.4143
## Kristendemokraterne -0.3926 -0.63083 -0.4505
## Liberal Alliance -0.3098 -0.54804 -0.3677
## Nye Borgerlige 0.0779 -0.16030 0.0200
## Radikale Venstre 0.1329 -0.10536 0.0750
## Socialistisk Folkeparti 0.3483 0.11008 0.2904
## Venstre -0.1807 -0.41892 -0.2386
## Vote blank 0.2412 0.00292 0.1832
## Would not vote 0.2016 -0.03660 0.1437
## Enhedslisten Konservative Folkeparti
## Socialdemokraterne 0.1434 -0.3564
## Alternativet -0.0948 -0.5946
## Dansk Folkeparti 0.0855 -0.4143
## Enhedslisten NA -0.4998
## Konservative Folkeparti -0.4998 NA
## Kristendemokraterne -0.5360 -0.0362
## Liberal Alliance -0.4532 0.0466
## Nye Borgerlige -0.0655 0.4343
## Radikale Venstre -0.0105 0.4893
## Socialistisk Folkeparti 0.2049 0.7047
## Venstre -0.3241 0.1757
## Vote blank 0.0978 0.5975
## Would not vote 0.0582 0.5580
## Kristendemokraterne Liberal Alliance Nye Borgerlige
## Socialdemokraterne -0.3926 -0.3098 0.0779
## Alternativet -0.6308 -0.5480 -0.1603
## Dansk Folkeparti -0.4505 -0.3677 0.0200
## Enhedslisten -0.5360 -0.4532 -0.0655
## Konservative Folkeparti -0.0362 0.0466 0.4343
## Kristendemokraterne NA 0.0828 0.4705
## Liberal Alliance 0.0828 NA 0.3877
## Nye Borgerlige 0.4705 0.3877 NA
## Radikale Venstre 0.5255 0.4427 0.0549
## Socialistisk Folkeparti 0.7409 0.6581 0.2704
## Venstre 0.2119 0.1291 -0.2586
## Vote blank 0.6337 0.5510 0.1632
## Would not vote 0.5942 0.5114 0.1237
## Radikale Venstre Socialistisk Folkeparti Venstre
## Socialdemokraterne 0.1329 0.348 -0.181
## Alternativet -0.1054 0.110 -0.419
## Dansk Folkeparti 0.0750 0.290 -0.239
## Enhedslisten -0.0105 0.205 -0.324
## Konservative Folkeparti 0.4893 0.705 0.176
## Kristendemokraterne 0.5255 0.741 0.212
## Liberal Alliance 0.4427 0.658 0.129
## Nye Borgerlige 0.0549 0.270 -0.259
## Radikale Venstre NA 0.215 -0.314
## Socialistisk Folkeparti 0.2154 NA -0.529
## Venstre -0.3136 -0.529 NA
## Vote blank 0.1083 -0.107 0.422
## Would not vote 0.0688 -0.147 0.382
## Vote blank Would not vote
## Socialdemokraterne 0.24116 0.2016
## Alternativet 0.00292 -0.0366
## Dansk Folkeparti 0.18325 0.1437
## Enhedslisten 0.09776 0.0582
## Konservative Folkeparti 0.59755 0.5580
## Kristendemokraterne 0.63375 0.5942
## Liberal Alliance 0.55095 0.5114
## Nye Borgerlige 0.16322 0.1237
## Radikale Venstre 0.10827 0.0688
## Socialistisk Folkeparti -0.10716 -0.1467
## Venstre 0.42183 0.3823
## Vote blank NA -0.0395
## Would not vote -0.03952 NASMD_matrix(d$Muslim_bias_r, d$party_vote)## Socialdemokraterne Alternativet Dansk Folkeparti
## Socialdemokraterne NA 0.2040 -0.2847
## Alternativet 0.2040 NA -0.4887
## Dansk Folkeparti -0.2847 -0.4887 NA
## Enhedslisten 0.1714 -0.0326 0.4561
## Konservative Folkeparti -0.5124 -0.7164 -0.2277
## Kristendemokraterne -0.5074 -0.7114 -0.2227
## Liberal Alliance -0.5391 -0.7431 -0.2544
## Nye Borgerlige -0.3147 -0.5187 -0.0300
## Radikale Venstre -0.0885 -0.2925 0.1962
## Socialistisk Folkeparti 0.1333 -0.0707 0.4180
## Venstre -0.2481 -0.4521 0.0365
## Vote blank 0.1639 -0.0402 0.4485
## Would not vote -0.1169 -0.3209 0.1678
## Enhedslisten Konservative Folkeparti
## Socialdemokraterne 0.17142 -0.51236
## Alternativet -0.03259 -0.71636
## Dansk Folkeparti 0.45609 -0.22769
## Enhedslisten NA -0.68378
## Konservative Folkeparti -0.68378 NA
## Kristendemokraterne -0.67883 0.00495
## Liberal Alliance -0.71049 -0.02671
## Nye Borgerlige -0.48608 0.19770
## Radikale Venstre -0.25991 0.42387
## Socialistisk Folkeparti -0.03809 0.64569
## Venstre -0.41955 0.26423
## Vote blank -0.00757 0.67621
## Would not vote -0.28828 0.39550
## Kristendemokraterne Liberal Alliance Nye Borgerlige
## Socialdemokraterne -0.50741 -0.5391 -0.3147
## Alternativet -0.71142 -0.7431 -0.5187
## Dansk Folkeparti -0.22274 -0.2544 -0.0300
## Enhedslisten -0.67883 -0.7105 -0.4861
## Konservative Folkeparti 0.00495 -0.0267 0.1977
## Kristendemokraterne NA -0.0317 0.1928
## Liberal Alliance -0.03166 NA 0.2244
## Nye Borgerlige 0.19275 0.2244 NA
## Radikale Venstre 0.41892 0.4506 0.2262
## Socialistisk Folkeparti 0.64074 0.6724 0.4480
## Venstre 0.25928 0.2909 0.0665
## Vote blank 0.67126 0.7029 0.4785
## Would not vote 0.39055 0.4222 0.1978
## Radikale Venstre Socialistisk Folkeparti Venstre
## Socialdemokraterne -0.0885 0.1333 -0.2481
## Alternativet -0.2925 -0.0707 -0.4521
## Dansk Folkeparti 0.1962 0.4180 0.0365
## Enhedslisten -0.2599 -0.0381 -0.4195
## Konservative Folkeparti 0.4239 0.6457 0.2642
## Kristendemokraterne 0.4189 0.6407 0.2593
## Liberal Alliance 0.4506 0.6724 0.2909
## Nye Borgerlige 0.2262 0.4480 0.0665
## Radikale Venstre NA 0.2218 -0.1596
## Socialistisk Folkeparti 0.2218 NA -0.3815
## Venstre -0.1596 -0.3815 NA
## Vote blank 0.2523 0.0305 0.4120
## Would not vote -0.0284 -0.2502 0.1313
## Vote blank Would not vote
## Socialdemokraterne 0.16385 -0.1169
## Alternativet -0.04015 -0.3209
## Dansk Folkeparti 0.44853 0.1678
## Enhedslisten -0.00757 -0.2883
## Konservative Folkeparti 0.67621 0.3955
## Kristendemokraterne 0.67126 0.3906
## Liberal Alliance 0.70292 0.4222
## Nye Borgerlige 0.47851 0.1978
## Radikale Venstre 0.25234 -0.0284
## Socialistisk Folkeparti 0.03052 -0.2502
## Venstre 0.41198 0.1313
## Vote blank NA -0.2807
## Would not vote -0.28071 NASMD_matrix(d$Muslim_preference, d$party_vote)## Socialdemokraterne Alternativet Dansk Folkeparti
## Socialdemokraterne NA -0.958 0.06673
## Alternativet -0.9581 NA 1.02484
## Dansk Folkeparti 0.0667 1.025 NA
## Enhedslisten -0.1505 0.808 -0.21727
## Konservative Folkeparti -0.2059 0.752 -0.27259
## Kristendemokraterne 0.2983 1.256 0.23153
## Liberal Alliance -0.2134 0.745 -0.28011
## Nye Borgerlige -0.0601 0.898 -0.12680
## Radikale Venstre -0.5574 0.401 -0.62417
## Socialistisk Folkeparti -0.0319 0.926 -0.09859
## Venstre 0.0690 1.027 0.00226
## Vote blank -0.0734 0.885 -0.14013
## Would not vote 0.0212 0.979 -0.04555
## Enhedslisten Konservative Folkeparti
## Socialdemokraterne -0.1505 -0.20587
## Alternativet 0.8076 0.75224
## Dansk Folkeparti -0.2173 -0.27259
## Enhedslisten NA -0.05533
## Konservative Folkeparti -0.0553 NA
## Kristendemokraterne 0.4488 0.50413
## Liberal Alliance -0.0628 -0.00752
## Nye Borgerlige 0.0905 0.14579
## Radikale Venstre -0.4069 -0.35157
## Socialistisk Folkeparti 0.1187 0.17401
## Venstre 0.2195 0.27485
## Vote blank 0.0771 0.13247
## Would not vote 0.1717 0.22704
## Kristendemokraterne Liberal Alliance Nye Borgerlige
## Socialdemokraterne 0.298 -0.21339 -0.0601
## Alternativet 1.256 0.74472 0.8980
## Dansk Folkeparti 0.232 -0.28011 -0.1268
## Enhedslisten 0.449 -0.06285 0.0905
## Konservative Folkeparti 0.504 -0.00752 0.1458
## Kristendemokraterne NA -0.51165 -0.3583
## Liberal Alliance -0.512 NA 0.1533
## Nye Borgerlige -0.358 0.15331 NA
## Radikale Venstre -0.856 -0.34405 -0.4974
## Socialistisk Folkeparti -0.330 0.18153 0.0282
## Venstre -0.229 0.28237 0.1291
## Vote blank -0.372 0.13999 -0.0133
## Would not vote -0.277 0.23456 0.0813
## Radikale Venstre Socialistisk Folkeparti Venstre
## Socialdemokraterne -0.557 -0.0319 0.06898
## Alternativet 0.401 0.9263 1.02709
## Dansk Folkeparti -0.624 -0.0986 0.00226
## Enhedslisten -0.407 0.1187 0.21952
## Konservative Folkeparti -0.352 0.1740 0.27485
## Kristendemokraterne -0.856 -0.3301 -0.22928
## Liberal Alliance -0.344 0.1815 0.28237
## Nye Borgerlige -0.497 0.0282 0.12906
## Radikale Venstre NA 0.5256 0.62642
## Socialistisk Folkeparti 0.526 NA 0.10084
## Venstre 0.626 0.1008 NA
## Vote blank 0.484 -0.0415 -0.14239
## Would not vote 0.579 0.0530 -0.04781
## Vote blank Would not vote
## Socialdemokraterne -0.0734 0.0212
## Alternativet 0.8847 0.9793
## Dansk Folkeparti -0.1401 -0.0456
## Enhedslisten 0.0771 0.1717
## Konservative Folkeparti 0.1325 0.2270
## Kristendemokraterne -0.3717 -0.2771
## Liberal Alliance 0.1400 0.2346
## Nye Borgerlige -0.0133 0.0813
## Radikale Venstre 0.4840 0.5786
## Socialistisk Folkeparti -0.0415 0.0530
## Venstre -0.1424 -0.0478
## Vote blank NA 0.0946
## Would not vote 0.0946 NASMD_matrix(d$stereotype_accuracy_noSYR, d$party_vote)## Socialdemokraterne Alternativet Dansk Folkeparti
## Socialdemokraterne NA 0.27664 0.0629
## Alternativet 0.2766 NA -0.2137
## Dansk Folkeparti 0.0629 -0.21371 NA
## Enhedslisten 0.1281 -0.14859 0.0651
## Konservative Folkeparti -0.1764 -0.45308 -0.2394
## Kristendemokraterne -0.4321 -0.70873 -0.4950
## Liberal Alliance -0.2628 -0.53946 -0.3258
## Nye Borgerlige 0.2823 0.00564 0.2194
## Radikale Venstre 0.1533 -0.12339 0.0903
## Socialistisk Folkeparti 0.4384 0.16177 0.3755
## Venstre -0.1747 -0.45134 -0.2376
## Vote blank 0.2994 0.02274 0.2365
## Would not vote 0.2879 0.01124 0.2249
## Enhedslisten Konservative Folkeparti
## Socialdemokraterne 0.1281 -0.17644
## Alternativet -0.1486 -0.45308
## Dansk Folkeparti 0.0651 -0.23937
## Enhedslisten NA -0.30450
## Konservative Folkeparti -0.3045 NA
## Kristendemokraterne -0.5601 -0.25565
## Liberal Alliance -0.3909 -0.08638
## Nye Borgerlige 0.1542 0.45872
## Radikale Venstre 0.0252 0.32969
## Socialistisk Folkeparti 0.3104 0.61485
## Venstre -0.3028 0.00174
## Vote blank 0.1713 0.47582
## Would not vote 0.1598 0.46432
## Kristendemokraterne Liberal Alliance Nye Borgerlige
## Socialdemokraterne -0.432 -0.2628 0.28228
## Alternativet -0.709 -0.5395 0.00564
## Dansk Folkeparti -0.495 -0.3258 0.21935
## Enhedslisten -0.560 -0.3909 0.15422
## Konservative Folkeparti -0.256 -0.0864 0.45872
## Kristendemokraterne NA 0.1693 0.71437
## Liberal Alliance 0.169 NA 0.54510
## Nye Borgerlige 0.714 0.5451 NA
## Radikale Venstre 0.585 0.4161 -0.12902
## Socialistisk Folkeparti 0.870 0.7012 0.15613
## Venstre 0.257 0.0881 -0.45698
## Vote blank 0.731 0.5622 0.01710
## Would not vote 0.720 0.5507 0.00560
## Radikale Venstre Socialistisk Folkeparti Venstre
## Socialdemokraterne 0.1533 0.438 -0.17470
## Alternativet -0.1234 0.162 -0.45134
## Dansk Folkeparti 0.0903 0.375 -0.23762
## Enhedslisten 0.0252 0.310 -0.30275
## Konservative Folkeparti 0.3297 0.615 0.00174
## Kristendemokraterne 0.5853 0.870 0.25739
## Liberal Alliance 0.4161 0.701 0.08813
## Nye Borgerlige -0.1290 0.156 -0.45698
## Radikale Venstre NA 0.285 -0.32795
## Socialistisk Folkeparti 0.2852 NA -0.61311
## Venstre -0.3280 -0.613 NA
## Vote blank 0.1461 -0.139 0.47408
## Would not vote 0.1346 -0.151 0.46257
## Vote blank Would not vote
## Socialdemokraterne 0.2994 0.2879
## Alternativet 0.0227 0.0112
## Dansk Folkeparti 0.2365 0.2249
## Enhedslisten 0.1713 0.1598
## Konservative Folkeparti 0.4758 0.4643
## Kristendemokraterne 0.7315 0.7200
## Liberal Alliance 0.5622 0.5507
## Nye Borgerlige 0.0171 0.0056
## Radikale Venstre 0.1461 0.1346
## Socialistisk Folkeparti -0.1390 -0.1505
## Venstre 0.4741 0.4626
## Vote blank NA -0.0115
## Would not vote -0.0115 NASMD_matrix(d$Muslim_bias_r_noSYR, d$party_vote)## Socialdemokraterne Alternativet Dansk Folkeparti
## Socialdemokraterne NA 0.2726 -0.2979
## Alternativet 0.273 NA -0.5705
## Dansk Folkeparti -0.298 -0.5705 NA
## Enhedslisten 0.193 -0.0796 0.4908
## Konservative Folkeparti -0.452 -0.7243 -0.1539
## Kristendemokraterne -0.461 -0.7333 -0.1629
## Liberal Alliance -0.531 -0.8035 -0.2330
## Nye Borgerlige -0.361 -0.6336 -0.0631
## Radikale Venstre -0.104 -0.3767 0.1937
## Socialistisk Folkeparti 0.186 -0.0869 0.4835
## Venstre -0.216 -0.4882 0.0823
## Vote blank 0.176 -0.0969 0.4735
## Would not vote -0.126 -0.3981 0.1723
## Enhedslisten Konservative Folkeparti
## Socialdemokraterne 0.19293 -0.45174
## Alternativet -0.07965 -0.72432
## Dansk Folkeparti 0.49081 -0.15386
## Enhedslisten NA -0.64467
## Konservative Folkeparti -0.64467 NA
## Kristendemokraterne -0.65369 -0.00902
## Liberal Alliance -0.72386 -0.07919
## Nye Borgerlige -0.55391 0.09076
## Radikale Venstre -0.29707 0.34760
## Socialistisk Folkeparti -0.00728 0.63739
## Venstre -0.40853 0.23614
## Vote blank -0.01727 0.62740
## Would not vote -0.31850 0.32617
## Kristendemokraterne Liberal Alliance Nye Borgerlige
## Socialdemokraterne -0.46076 -0.5309 -0.3610
## Alternativet -0.73334 -0.8035 -0.6336
## Dansk Folkeparti -0.16288 -0.2330 -0.0631
## Enhedslisten -0.65369 -0.7239 -0.5539
## Konservative Folkeparti -0.00902 -0.0792 0.0908
## Kristendemokraterne NA -0.0702 0.0998
## Liberal Alliance -0.07017 NA 0.1699
## Nye Borgerlige 0.09978 0.1699 NA
## Radikale Venstre 0.35662 0.4268 0.2568
## Socialistisk Folkeparti 0.64641 0.7166 0.5466
## Venstre 0.24516 0.3153 0.1454
## Vote blank 0.63642 0.7066 0.5366
## Would not vote 0.33519 0.4054 0.2354
## Radikale Venstre Socialistisk Folkeparti Venstre
## Socialdemokraterne -0.1041 0.18565 -0.2156
## Alternativet -0.3767 -0.08693 -0.4882
## Dansk Folkeparti 0.1937 0.48353 0.0823
## Enhedslisten -0.2971 -0.00728 -0.4085
## Konservative Folkeparti 0.3476 0.63739 0.2361
## Kristendemokraterne 0.3566 0.64641 0.2452
## Liberal Alliance 0.4268 0.71657 0.3153
## Nye Borgerlige 0.2568 0.54663 0.1454
## Radikale Venstre NA 0.28979 -0.1115
## Socialistisk Folkeparti 0.2898 NA -0.4013
## Venstre -0.1115 -0.40125 NA
## Vote blank 0.2798 -0.00999 0.3913
## Would not vote -0.0214 -0.31121 0.0900
## Vote blank Would not vote
## Socialdemokraterne 0.17566 -0.1256
## Alternativet -0.09692 -0.3981
## Dansk Folkeparti 0.47354 0.1723
## Enhedslisten -0.01727 -0.3185
## Konservative Folkeparti 0.62740 0.3262
## Kristendemokraterne 0.63642 0.3352
## Liberal Alliance 0.70658 0.4054
## Nye Borgerlige 0.53664 0.2354
## Radikale Venstre 0.27980 -0.0214
## Socialistisk Folkeparti -0.00999 -0.3112
## Venstre 0.39126 0.0900
## Vote blank NA -0.3012
## Would not vote -0.30122 NASMD_matrix(d$Muslim_preference_noSYR, d$party_vote)## Socialdemokraterne Alternativet Dansk Folkeparti
## Socialdemokraterne NA -0.976 -0.00643
## Alternativet -0.97608 NA 0.96965
## Dansk Folkeparti -0.00643 0.970 NA
## Enhedslisten -0.20590 0.770 -0.19947
## Konservative Folkeparti -0.27258 0.704 -0.26615
## Kristendemokraterne 0.20832 1.184 0.21475
## Liberal Alliance -0.29318 0.683 -0.28675
## Nye Borgerlige -0.11862 0.857 -0.11219
## Radikale Venstre -0.62852 0.348 -0.62209
## Socialistisk Folkeparti -0.05726 0.919 -0.05083
## Venstre -0.00883 0.967 -0.00240
## Vote blank -0.16802 0.808 -0.16159
## Would not vote -0.06955 0.907 -0.06312
## Enhedslisten Konservative Folkeparti
## Socialdemokraterne -0.2059 -0.2726
## Alternativet 0.7702 0.7035
## Dansk Folkeparti -0.1995 -0.2661
## Enhedslisten NA -0.0667
## Konservative Folkeparti -0.0667 NA
## Kristendemokraterne 0.4142 0.4809
## Liberal Alliance -0.0873 -0.0206
## Nye Borgerlige 0.0873 0.1540
## Radikale Venstre -0.4226 -0.3559
## Socialistisk Folkeparti 0.1486 0.2153
## Venstre 0.1971 0.2637
## Vote blank 0.0379 0.1046
## Would not vote 0.1363 0.2030
## Kristendemokraterne Liberal Alliance Nye Borgerlige
## Socialdemokraterne 0.208 -0.2932 -0.1186
## Alternativet 1.184 0.6829 0.8575
## Dansk Folkeparti 0.215 -0.2867 -0.1122
## Enhedslisten 0.414 -0.0873 0.0873
## Konservative Folkeparti 0.481 -0.0206 0.1540
## Kristendemokraterne NA -0.5015 -0.3269
## Liberal Alliance -0.501 NA 0.1746
## Nye Borgerlige -0.327 0.1746 NA
## Radikale Venstre -0.837 -0.3353 -0.5099
## Socialistisk Folkeparti -0.266 0.2359 0.0614
## Venstre -0.217 0.2843 0.1098
## Vote blank -0.376 0.1252 -0.0494
## Would not vote -0.278 0.2236 0.0491
## Radikale Venstre Socialistisk Folkeparti Venstre
## Socialdemokraterne -0.629 -0.0573 -0.00883
## Alternativet 0.348 0.9188 0.96724
## Dansk Folkeparti -0.622 -0.0508 -0.00240
## Enhedslisten -0.423 0.1486 0.19707
## Konservative Folkeparti -0.356 0.2153 0.26374
## Kristendemokraterne -0.837 -0.2656 -0.21715
## Liberal Alliance -0.335 0.2359 0.28434
## Nye Borgerlige -0.510 0.0614 0.10979
## Radikale Venstre NA 0.5713 0.61969
## Socialistisk Folkeparti 0.571 NA 0.04842
## Venstre 0.620 0.0484 NA
## Vote blank 0.461 -0.1108 -0.15918
## Would not vote 0.559 -0.0123 -0.06072
## Vote blank Would not vote
## Socialdemokraterne -0.1680 -0.0696
## Alternativet 0.8081 0.9065
## Dansk Folkeparti -0.1616 -0.0631
## Enhedslisten 0.0379 0.1363
## Konservative Folkeparti 0.1046 0.2030
## Kristendemokraterne -0.3763 -0.2779
## Liberal Alliance 0.1252 0.2236
## Nye Borgerlige -0.0494 0.0491
## Radikale Venstre 0.4605 0.5590
## Socialistisk Folkeparti -0.1108 -0.0123
## Venstre -0.1592 -0.0607
## Vote blank NA 0.0985
## Would not vote 0.0985 NA#simplified coding
#with SYR
rms::ols(stereotype_accuracy ~ block_vote, data = d)## Linear Regression Model
##
## rms::ols(formula = stereotype_accuracy ~ block_vote, data = d)
##
## Model Likelihood Discrimination
## Ratio Test Indexes
## Obs 476 LR chi2 7.59 R2 0.016
## sigma0.1977 d.f. 3 R2 adj 0.010
## d.f. 472 Pr(> chi2) 0.0553 g 0.026
##
## Residuals
##
## Min 1Q Median 3Q Max
## -1.05708 -0.06121 0.04658 0.12682 0.31228
##
##
## Coef S.E. t Pr(>|t|)
## Intercept 0.6064 0.0139 43.70 <0.0001
## block_vote=left-block -0.0441 0.0196 -2.25 0.0246
## block_vote=Vote blank -0.0656 0.0326 -2.01 0.0446
## block_vote=Would not vote -0.0578 0.0435 -1.33 0.1846
## rms::ols(Muslim_bias_r ~ block_vote, data = d)## Linear Regression Model
##
## rms::ols(formula = Muslim_bias_r ~ block_vote, data = d)
##
## Model Likelihood Discrimination
## Ratio Test Indexes
## Obs 476 LR chi2 19.86 R2 0.041
## sigma0.1579 d.f. 3 R2 adj 0.035
## d.f. 472 Pr(> chi2) 0.0002 g 0.034
##
## Residuals
##
## Min 1Q Median 3Q Max
## -0.29360 -0.12292 -0.01088 0.10494 0.48637
##
##
## Coef S.E. t Pr(>|t|)
## Intercept -0.4495 0.0111 -40.57 <0.0001
## block_vote=left-block -0.0635 0.0156 -4.06 <0.0001
## block_vote=Vote blank -0.0789 0.0260 -3.03 0.0026
## block_vote=Would not vote -0.0344 0.0347 -0.99 0.3230
## rms::ols(Muslim_preference ~ block_vote, data = d)## Frequencies of Missing Values Due to Each Variable
## Muslim_preference block_vote
## 108 0
##
## Linear Regression Model
##
## rms::ols(formula = Muslim_preference ~ block_vote, data = d)
##
##
## Model Likelihood Discrimination
## Ratio Test Indexes
## Obs 368 LR chi2 4.81 R2 0.013
## sigma0.4224 d.f. 3 R2 adj 0.005
## d.f. 364 Pr(> chi2) 0.1860 g 0.050
##
## Residuals
##
## Min 1Q Median 3Q Max
## -0.84521 -0.29235 -0.04535 0.27061 1.26140
##
##
## Coef S.E. t Pr(>|t|)
## Intercept -0.2551 0.0329 -7.76 <0.0001
## block_vote=left-block 0.1003 0.0472 2.12 0.0344
## block_vote=Vote blank 0.0331 0.0816 0.41 0.6854
## block_vote=Would not vote -0.0063 0.1106 -0.06 0.9546
## #without
rms::ols(stereotype_accuracy_noSYR ~ block_vote, data = d)## Frequencies of Missing Values Due to Each Variable
## stereotype_accuracy_noSYR block_vote
## 2 0
##
## Linear Regression Model
##
## rms::ols(formula = stereotype_accuracy_noSYR ~ block_vote, data = d)
##
##
## Model Likelihood Discrimination
## Ratio Test Indexes
## Obs 474 LR chi2 7.69 R2 0.016
## sigma0.2079 d.f. 3 R2 adj 0.010
## d.f. 470 Pr(> chi2) 0.0528 g 0.028
##
## Residuals
##
## Min 1Q Median 3Q Max
## -1.16439 -0.06770 0.05161 0.12999 0.28290
##
##
## Coef S.E. t Pr(>|t|)
## Intercept 0.6268 0.0146 42.85 <0.0001
## block_vote=left-block -0.0423 0.0206 -2.05 0.0408
## block_vote=Vote blank -0.0729 0.0343 -2.13 0.0339
## block_vote=Would not vote -0.0705 0.0458 -1.54 0.1239
## rms::ols(Muslim_bias_r_noSYR ~ block_vote, data = d)## Linear Regression Model
##
## rms::ols(formula = Muslim_bias_r_noSYR ~ block_vote, data = d)
##
## Model Likelihood Discrimination
## Ratio Test Indexes
## Obs 476 LR chi2 21.06 R2 0.043
## sigma0.1590 d.f. 3 R2 adj 0.037
## d.f. 472 Pr(> chi2) 0.0001 g 0.035
##
## Residuals
##
## Min 1Q Median 3Q Max
## -0.29752 -0.12214 -0.01422 0.09913 0.54025
##
##
## Coef S.E. t Pr(>|t|)
## Intercept -0.4902 0.0112 -43.92 <0.0001
## block_vote=left-block -0.0663 0.0157 -4.21 <0.0001
## block_vote=Vote blank -0.0806 0.0262 -3.08 0.0022
## block_vote=Would not vote -0.0325 0.0350 -0.93 0.3530
## rms::ols(Muslim_preference_noSYR ~ block_vote, data = d)## Frequencies of Missing Values Due to Each Variable
## Muslim_preference_noSYR block_vote
## 114 0
##
## Linear Regression Model
##
## rms::ols(formula = Muslim_preference_noSYR ~ block_vote, data = d)
##
##
## Model Likelihood Discrimination
## Ratio Test Indexes
## Obs 362 LR chi2 2.89 R2 0.008
## sigma0.4244 d.f. 3 R2 adj 0.000
## d.f. 358 Pr(> chi2) 0.4094 g 0.039
##
## Residuals
##
## Min 1Q Median 3Q Max
## -0.82491 -0.29259 -0.04352 0.25378 1.25456
##
##
## Coef S.E. t Pr(>|t|)
## Intercept -0.2546 0.0331 -7.68 <0.0001
## block_vote=left-block 0.0795 0.0480 1.66 0.0984
## block_vote=Vote blank 0.0424 0.0820 0.52 0.6052
## block_vote=Would not vote 0.0012 0.1112 0.01 0.9911
## #cohen d
cohen.d(d %>% select(stereotype_accuracy, stereotype_accuracy_noSYR, Muslim_bias_r, Muslim_bias_r_noSYR, Muslim_preference, Muslim_preference_noSYR, block_vote), "block_vote") %>%
{
cat("p values for gaps:\n")
print(.$p)
.
}## p values for gaps:
## stereotype_accuracy stereotype_accuracy_noSYR Muslim_bias_r
## 2.44e-02 3.96e-02 5.54e-05
## Muslim_bias_r_noSYR Muslim_preference Muslim_preference_noSYR
## 3.21e-05 3.16e-02 9.35e-02## Call: cohen.d(x = d %>% select(stereotype_accuracy, stereotype_accuracy_noSYR,
## Muslim_bias_r, Muslim_bias_r_noSYR, Muslim_preference, Muslim_preference_noSYR,
## block_vote), group = "block_vote")
## Cohen d statistic of difference between two means
## lower effect upper
## stereotype_accuracy -0.42 -0.22 -0.03
## stereotype_accuracy_noSYR -0.40 -0.20 -0.01
## Muslim_bias_r -0.60 -0.40 -0.20
## Muslim_bias_r_noSYR -0.61 -0.42 -0.22
## Muslim_preference 0.02 0.24 0.46
## Muslim_preference_noSYR -0.03 0.19 0.41
##
## Multivariate (Mahalanobis) distance between groups
## [1] 0.65
## r equivalent of difference between two means
## stereotype_accuracy stereotype_accuracy_noSYR Muslim_bias_r
## -0.11 -0.10 -0.20
## Muslim_bias_r_noSYR Muslim_preference Muslim_preference_noSYR
## -0.20 0.12 0.09#plot party means
GG_group_means(d, "stereotype_accuracy", "party_vote") +
theme(axis.text.x = element_text(angle = -80, hjust = 0)) +
scale_y_continuous("Stereotype accuracy\n(pearson r)") +
scale_x_discrete("Intended party")## Scale for 'x' is already present. Adding another scale for 'x', which will
## replace the existing scale.
GG_save("figures/accuracy_by_party.png")
GG_group_means(d %>% df_standardize(messages = F), "stereotype_accuracy", "party_vote") +
theme(axis.text.x = element_text(angle = -80, hjust = 0)) +
scale_y_continuous("Stereotype accuracy\n(pearson r, standardized)") +
scale_x_discrete("Intended party")## Scale for 'x' is already present. Adding another scale for 'x', which will
## replace the existing scale.
GG_save("figures/accuracy_by_party_standardized.png")
GG_group_means(d, "Muslim_bias_r", "party_vote") +
theme(axis.text.x = element_text(angle = -80, hjust = 0)) +
scale_y_continuous("Muslim bias in steroetype") +
scale_x_discrete("Intended party")## Scale for 'x' is already present. Adding another scale for 'x', which will
## replace the existing scale.
GG_save("figures/Muslim_bias_by_party.png")
GG_group_means(d %>% df_standardize(messages = F), "Muslim_bias_r", "party_vote") +
theme(axis.text.x = element_text(angle = -80, hjust = 0)) +
scale_y_continuous("Muslim bias in steroetype (standardized)") +
scale_x_discrete("Intended party")## Scale for 'x' is already present. Adding another scale for 'x', which will
## replace the existing scale.
GG_save("figures/Muslim_bias_by_party_standardized.png")
GG_group_means(d, "Muslim_preference", "party_vote") +
theme(axis.text.x = element_text(angle = -80, hjust = 0)) +
scale_y_continuous("Muslim preference") +
scale_x_discrete("Intended party")## Missing values were removed.## Scale for 'x' is already present. Adding another scale for 'x', which will
## replace the existing scale.
GG_save("figures/Muslim_preference_by_party.png")
GG_group_means(d %>% df_standardize(messages = F), "Muslim_preference", "party_vote") +
theme(axis.text.x = element_text(angle = -80, hjust = 0)) +
scale_y_continuous("Muslim preference (standardized)") +
scale_x_discrete("Intended party")## Missing values were removed.
## Scale for 'x' is already present. Adding another scale for 'x', which will
## replace the existing scale.
GG_save("figures/Muslim_preference_by_party_standardized.png")
#no SYR
GG_group_means(d, "stereotype_accuracy_noSYR", "party_vote") +
theme(axis.text.x = element_text(angle = -80, hjust = 0)) +
scale_y_continuous("Stereotype accuracy\n(pearson r)") +
scale_x_discrete("Intended party") +
labs(caption = "Syria excluded")## Missing values were removed.
## Scale for 'x' is already present. Adding another scale for 'x', which will
## replace the existing scale.
GG_save("figures/accuracy_by_party_noSYR.png")
GG_group_means(d %>% df_standardize(messages = F), "stereotype_accuracy_noSYR", "party_vote") +
theme(axis.text.x = element_text(angle = -80, hjust = 0)) +
scale_y_continuous("Stereotype accuracy\n(pearson r, standardized)") +
scale_x_discrete("Intended party") +
labs(caption = "Syria excluded")## Missing values were removed.
## Scale for 'x' is already present. Adding another scale for 'x', which will
## replace the existing scale.
GG_save("figures/accuracy_by_party_noSYR_standardized.png")
GG_group_means(d, "Muslim_bias_r_noSYR", "party_vote") +
theme(axis.text.x = element_text(angle = -80, hjust = 0)) +
scale_y_continuous("Muslim bias in steroetype") +
scale_x_discrete("Intended party") +
labs(caption = "Syria excluded")## Scale for 'x' is already present. Adding another scale for 'x', which will
## replace the existing scale.
GG_save("figures/Muslim_bias_by_party_noSYR.png")
GG_group_means(d %>% df_standardize(messages = F), "Muslim_bias_r_noSYR", "party_vote") +
theme(axis.text.x = element_text(angle = -80, hjust = 0)) +
scale_y_continuous("Muslim bias in steroetype (standardized)") +
scale_x_discrete("Intended party") +
labs(caption = "Syria excluded")## Scale for 'x' is already present. Adding another scale for 'x', which will
## replace the existing scale.
GG_save("figures/Muslim_bias_by_party_noSYR_standardized.png")
GG_group_means(d, "Muslim_preference_noSYR", "party_vote") +
theme(axis.text.x = element_text(angle = -80, hjust = 0)) +
scale_y_continuous("Muslim preference") +
scale_x_discrete("Intended party") +
labs(caption = "Syria excluded")## Missing values were removed.
## Scale for 'x' is already present. Adding another scale for 'x', which will
## replace the existing scale.
GG_save("figures/Muslim_preference_by_party_noSYR.png")
GG_group_means(d %>% df_standardize(messages = F), "Muslim_preference_noSYR", "party_vote") +
theme(axis.text.x = element_text(angle = -80, hjust = 0)) +
scale_y_continuous("Muslim preference (standardized)") +
scale_x_discrete("Intended party") +
labs(caption = "Syria excluded")## Missing values were removed.
## Scale for 'x' is already present. Adding another scale for 'x', which will
## replace the existing scale.
GG_save("figures/Muslim_preference_by_party_noSYR_standardized.png")Order of presentation
#order of presentation
rms::ols(stereotype_accuracy ~ path_parsed, data = d)## Linear Regression Model
##
## rms::ols(formula = stereotype_accuracy ~ path_parsed, data = d)
##
## Model Likelihood Discrimination
## Ratio Test Indexes
## Obs 476 LR chi2 2.64 R2 0.006
## sigma0.1992 d.f. 5 R2 adj -0.005
## d.f. 470 Pr(> chi2) 0.7547 g 0.017
##
## Residuals
##
## Min 1Q Median 3Q Max
## -1.02251 -0.06574 0.04677 0.12806 0.28813
##
##
## Coef S.E. t Pr(>|t|)
## Intercept 0.5857 0.0224 26.14 <0.0001
## path_parsed=ACB 0.0151 0.0317 0.48 0.6334
## path_parsed=BAC -0.0304 0.0310 -0.98 0.3279
## path_parsed=BCA -0.0081 0.0305 -0.27 0.7900
## path_parsed=CAB 0.0006 0.0332 0.02 0.9846
## path_parsed=CBA -0.0189 0.0323 -0.58 0.5594
## rms::ols(Muslim_bias_r ~ path_parsed, data = d)## Linear Regression Model
##
## rms::ols(formula = Muslim_bias_r ~ path_parsed, data = d)
##
## Model Likelihood Discrimination
## Ratio Test Indexes
## Obs 476 LR chi2 5.55 R2 0.012
## sigma0.1606 d.f. 5 R2 adj 0.001
## d.f. 470 Pr(> chi2) 0.3521 g 0.019
##
## Residuals
##
## Min 1Q Median 3Q Max
## -0.27694 -0.12006 -0.01166 0.10619 0.50347
##
##
## Coef S.E. t Pr(>|t|)
## Intercept -0.4852 0.0181 -26.85 <0.0001
## path_parsed=ACB 0.0205 0.0256 0.80 0.4235
## path_parsed=BAC -0.0097 0.0250 -0.39 0.6978
## path_parsed=BCA 0.0061 0.0246 0.25 0.8029
## path_parsed=CAB 0.0125 0.0268 0.47 0.6410
## path_parsed=CBA -0.0346 0.0261 -1.33 0.1849
## rms::ols(Muslim_preference ~ path_parsed, data = d)## Frequencies of Missing Values Due to Each Variable
## Muslim_preference path_parsed
## 108 0
##
## Linear Regression Model
##
## rms::ols(formula = Muslim_preference ~ path_parsed, data = d)
##
##
## Model Likelihood Discrimination
## Ratio Test Indexes
## Obs 368 LR chi2 3.91 R2 0.011
## sigma0.4241 d.f. 5 R2 adj -0.003
## d.f. 362 Pr(> chi2) 0.5618 g 0.047
##
## Residuals
##
## Min 1Q Median 3Q Max
## -0.81406 -0.29094 -0.05197 0.27470 1.22128
##
##
## Coef S.E. t Pr(>|t|)
## Intercept -0.2213 0.0562 -3.94 <0.0001
## path_parsed=ACB -0.0398 0.0772 -0.52 0.6063
## path_parsed=BAC -0.0242 0.0764 -0.32 0.7514
## path_parsed=BCA 0.0353 0.0741 0.48 0.6337
## path_parsed=CAB 0.1037 0.0841 1.23 0.2183
## path_parsed=CBA 0.0130 0.0794 0.16 0.8699
## rms::ols(stereotype_accuracy_noSYR ~ path_parsed, data = d)## Frequencies of Missing Values Due to Each Variable
## stereotype_accuracy_noSYR path_parsed
## 2 0
##
## Linear Regression Model
##
## rms::ols(formula = stereotype_accuracy_noSYR ~ path_parsed, data = d)
##
##
## Model Likelihood Discrimination
## Ratio Test Indexes
## Obs 474 LR chi2 1.29 R2 0.003
## sigma0.2098 d.f. 5 R2 adj -0.008
## d.f. 468 Pr(> chi2) 0.9355 g 0.012
##
## Residuals
##
## Min 1Q Median 3Q Max
## -1.12539 -0.06203 0.05790 0.13341 0.28529
##
##
## Coef S.E. t Pr(>|t|)
## Intercept 0.6126 0.0236 25.96 <0.0001
## path_parsed=ACB -0.0022 0.0334 -0.07 0.9465
## path_parsed=BAC -0.0248 0.0327 -0.76 0.4481
## path_parsed=BCA -0.0281 0.0323 -0.87 0.3836
## path_parsed=CAB -0.0103 0.0350 -0.29 0.7694
## path_parsed=CBA -0.0174 0.0341 -0.51 0.6106
## rms::ols(Muslim_bias_r_noSYR ~ path_parsed, data = d)## Linear Regression Model
##
## rms::ols(formula = Muslim_bias_r_noSYR ~ path_parsed, data = d)
##
## Model Likelihood Discrimination
## Ratio Test Indexes
## Obs 476 LR chi2 4.96 R2 0.010
## sigma0.1621 d.f. 5 R2 adj 0.000
## d.f. 470 Pr(> chi2) 0.4208 g 0.017
##
## Residuals
##
## Min 1Q Median 3Q Max
## -0.28542 -0.11790 -0.01894 0.09918 0.53380
##
##
## Coef S.E. t Pr(>|t|)
## Intercept -0.5263 0.0182 -28.86 <0.0001
## path_parsed=ACB 0.0151 0.0258 0.59 0.5578
## path_parsed=BAC -0.0087 0.0253 -0.34 0.7316
## path_parsed=BCA 0.0065 0.0248 0.26 0.7926
## path_parsed=CAB 0.0116 0.0270 0.43 0.6673
## path_parsed=CBA -0.0357 0.0263 -1.36 0.1752
## rms::ols(Muslim_preference_noSYR ~ path_parsed, data = d)## Frequencies of Missing Values Due to Each Variable
## Muslim_preference_noSYR path_parsed
## 114 0
##
## Linear Regression Model
##
## rms::ols(formula = Muslim_preference_noSYR ~ path_parsed, data = d)
##
##
## Model Likelihood Discrimination
## Ratio Test Indexes
## Obs 362 LR chi2 4.91 R2 0.013
## sigma0.4244 d.f. 5 R2 adj 0.000
## d.f. 356 Pr(> chi2) 0.4268 g 0.053
##
## Residuals
##
## Min 1Q Median 3Q Max
## -0.81910 -0.28468 -0.04521 0.24950 1.26611
##
##
## Coef S.E. t Pr(>|t|)
## Intercept -0.2363 0.0562 -4.20 <0.0001
## path_parsed=ACB -0.0264 0.0773 -0.34 0.7327
## path_parsed=BAC -0.0298 0.0773 -0.39 0.7004
## path_parsed=BCA 0.0554 0.0744 0.75 0.4564
## path_parsed=CAB 0.1209 0.0846 1.43 0.1540
## path_parsed=CBA 0.0115 0.0799 0.14 0.8854
## Reviewer requests
#summary stats of estimates by country
d_econ_num_sumstats = d_econ_num %>% map_df(function(dd) {
describe(dd) %>%
as.data.frame() %>%
mutate(
prop_1 = mean(dd == 1),
prop_7 = mean(dd == 7),
)
}) %>%
select(mean, median, sd, mad, skew, kurtosis, prop_1, prop_7) %>%
df_add_affix(prefix = "est_")
#summary stats of preferences by country
d_prefs_num_sumstats = d_prefs_num_alt %>% map_df(function(dd) {
describe(dd) %>%
as.data.frame() %>%
mutate(
prop_1 = mean(dd == 1),
prop_4 = mean(dd == 4),
)
}) %>%
select(mean, median, sd, mad, skew, kurtosis, prop_1, prop_4) %>%
df_add_affix(prefix = "pref_")
#split age into 3 groups
d$age_bin = d$age %>% discretize(3, equal_range = F)
#join with data
d_reviewer = cbind(
#estimates
d_econ_num_sumstats,
est_mean_young = colMeans(d_econ_num[d$age_bin == "[18,31]", ]),
est_mean_middle = colMeans(d_econ_num[d$age_bin == "(31,46]", ]),
est_mean_old = colMeans(d_econ_num[d$age_bin == "(46,65]", ]),
#prefs
d_prefs_num_sumstats,
pref_mean_young = colMeans(d_prefs_num[d$age_bin == "[18,31]", ]),
pref_mean_middle = colMeans(d_prefs_num[d$age_bin == "(31,46]", ]),
pref_mean_old = colMeans(d_prefs_num[d$age_bin == "(46,65]", ]),
#covariates
muslim = dk_fiscal_sub$Muslim,
net_fiscal = dk_fiscal_sub$dnk_net_fiscal_contribution_2014
)
#nice print
d_reviewer %>% df_round() %>% DT::datatable()#estimates by age group in plot
d_reviewer %>%
select(est_mean_young:est_mean_old) %>%
mutate(name = dk_fiscal_sub$Names) %>%
gather(key = age_group, value = mean, -name) %>%
mutate(age_group = ordered(age_group, levels = c("est_mean_young", "est_mean_middle", "est_mean_old"))) %>%
ggplot(aes(name, mean, fill = age_group)) +
geom_bar(stat = "identity", position = "dodge") +
scale_x_discrete(NULL, guide = guide_axis(angle = -28)) +
ylab("Mean estimate of net fiscal contribution") +
scale_fill_discrete("Age group", labels = levels(d$age_bin))
GG_save("figures/age_groups_estimates.png")
#preferences by age group in plot
d_reviewer %>%
select(pref_mean_young:pref_mean_old) %>%
mutate(name = dk_fiscal_sub$Names) %>%
gather(key = age_group, value = mean, -name) %>%
mutate(age_group = ordered(age_group, levels = c("pref_mean_young", "pref_mean_middle", "pref_mean_old"))) %>%
ggplot(aes(name, mean, fill = age_group)) +
geom_bar(stat = "identity", position = "dodge") +
scale_x_discrete(NULL, guide = guide_axis(angle = -28)) +
ylab("Net opposition") +
scale_fill_discrete("Age group", labels = levels(d$age_bin))
GG_save("figures/age_groups_preferences.png")
#summary stats
d_reviewer %>%
select(est_mean_young:est_mean_old, pref_mean_young:pref_mean_old) %>%
describe()## vars n mean sd median trimmed mad min max range
## est_mean_young 1 32 3.8469 1.026 3.7000 3.8514 1.274 2.087 5.450 3.36
## est_mean_middle 2 32 3.9407 1.052 3.8690 3.9675 1.377 1.982 5.530 3.55
## est_mean_old 3 32 3.7403 1.231 3.5372 3.7256 1.538 1.696 5.696 4.00
## pref_mean_young 4 32 -0.2734 0.350 -0.2125 -0.2755 0.389 -0.800 0.263 1.06
## pref_mean_middle 5 32 -0.1581 0.481 -0.0536 -0.1525 0.556 -0.857 0.607 1.46
## pref_mean_old 6 32 0.0173 0.556 0.2095 0.0385 0.551 -0.838 0.730 1.57
## skew kurtosis se
## est_mean_young 0.1121 -1.22 0.1813
## est_mean_middle -0.0424 -1.19 0.1859
## est_mean_old 0.2138 -1.30 0.2177
## pref_mean_young -0.0832 -1.39 0.0619
## pref_mean_middle -0.2295 -1.43 0.0850
## pref_mean_old -0.4061 -1.49 0.0983#cors
d_reviewer %>%
wtd.cors()## est_mean est_median est_sd est_mad est_skew est_kurtosis
## est_mean 1.000 0.970 -0.329787 Inf -0.976 0.641
## est_median 0.970 1.000 -0.288226 Inf -0.948 0.631
## est_sd -0.330 -0.288 1.000000 Inf 0.299 -0.723
## est_mad Inf Inf Inf NaN Inf Inf
## est_skew -0.976 -0.948 0.298506 Inf 1.000 -0.606
## est_kurtosis 0.641 0.631 -0.722663 Inf -0.606 1.000
## est_prop_1 -0.862 -0.845 -0.029734 Inf 0.880 -0.210
## est_prop_7 0.859 0.847 -0.455670 Inf -0.826 0.838
## est_mean_young 0.995 0.969 -0.351903 -Inf -0.980 0.650
## est_mean_middle 0.997 0.968 -0.277612 Inf -0.976 0.599
## est_mean_old 0.996 0.963 -0.356958 Inf -0.961 0.668
## pref_mean 0.987 0.954 -0.410820 Inf -0.945 0.689
## pref_median 0.799 0.768 -0.225471 Inf -0.738 0.489
## pref_sd -0.953 -0.892 0.429427 NaN 0.915 -0.627
## pref_mad -0.845 -0.771 0.359725 -Inf 0.774 -0.602
## pref_skew -0.958 -0.928 0.229719 NaN 0.932 -0.503
## pref_kurtosis 0.883 0.818 -0.633750 -Inf -0.812 0.827
## pref_prop_1 -0.960 -0.930 0.189298 Inf 0.942 -0.460
## pref_prop_4 0.856 0.830 -0.609950 Inf -0.804 0.880
## pref_mean_young -0.967 -0.934 0.354461 NaN 0.936 -0.618
## pref_mean_middle -0.980 -0.948 0.409438 -Inf 0.931 -0.673
## pref_mean_old -0.970 -0.931 0.468535 Inf 0.915 -0.731
## muslim -0.723 -0.710 0.000769 -Inf 0.703 -0.260
## net_fiscal 0.806 0.775 -0.040950 -Inf -0.804 0.310
## est_prop_1 est_prop_7 est_mean_young est_mean_middle
## est_mean -0.8623 0.859 0.995 0.997
## est_median -0.8449 0.847 0.969 0.968
## est_sd -0.0297 -0.456 -0.352 -0.278
## est_mad Inf Inf -Inf Inf
## est_skew 0.8801 -0.826 -0.980 -0.976
## est_kurtosis -0.2098 0.838 0.650 0.599
## est_prop_1 1.0000 -0.555 -0.857 -0.885
## est_prop_7 -0.5555 1.000 0.855 0.839
## est_mean_young -0.8565 0.855 1.000 0.990
## est_mean_middle -0.8854 0.839 0.990 1.000
## est_mean_old -0.8359 0.872 0.985 0.991
## pref_mean -0.8168 0.868 0.981 0.980
## pref_median -0.6663 0.639 0.789 0.799
## pref_sd 0.8050 -0.807 -0.948 -0.948
## pref_mad 0.6524 -0.718 -0.835 -0.835
## pref_skew 0.9104 -0.738 -0.952 -0.965
## pref_kurtosis -0.5753 0.872 0.878 0.859
## pref_prop_1 0.9430 -0.723 -0.952 -0.970
## pref_prop_4 -0.5206 0.946 0.855 0.829
## pref_mean_young 0.8391 -0.807 -0.969 -0.965
## pref_mean_middle 0.8153 -0.843 -0.974 -0.974
## pref_mean_old 0.7587 -0.870 -0.961 -0.958
## muslim 0.7690 -0.479 -0.717 -0.748
## net_fiscal -0.8342 0.572 0.815 0.818
## est_mean_old pref_mean pref_median pref_sd pref_mad pref_skew
## est_mean 0.996 0.987 0.799 -0.953 -0.845 -0.958
## est_median 0.963 0.954 0.768 -0.892 -0.771 -0.928
## est_sd -0.357 -0.411 -0.225 0.429 0.360 0.230
## est_mad Inf Inf Inf NaN -Inf NaN
## est_skew -0.961 -0.945 -0.738 0.915 0.774 0.932
## est_kurtosis 0.668 0.689 0.489 -0.627 -0.602 -0.503
## est_prop_1 -0.836 -0.817 -0.666 0.805 0.652 0.910
## est_prop_7 0.872 0.868 0.639 -0.807 -0.718 -0.738
## est_mean_young 0.985 0.981 0.789 -0.948 -0.835 -0.952
## est_mean_middle 0.991 0.980 0.799 -0.948 -0.835 -0.965
## est_mean_old 1.000 0.988 0.800 -0.953 -0.855 -0.947
## pref_mean 0.988 1.000 0.826 -0.960 -0.865 -0.957
## pref_median 0.800 0.826 1.000 -0.780 -0.882 -0.860
## pref_sd -0.953 -0.960 -0.780 1.000 0.836 0.928
## pref_mad -0.855 -0.865 -0.882 0.836 1.000 0.863
## pref_skew -0.947 -0.957 -0.860 0.928 0.863 1.000
## pref_kurtosis 0.902 0.927 0.740 -0.913 -0.835 -0.813
## pref_prop_1 -0.947 -0.946 -0.790 0.929 0.794 0.978
## pref_prop_4 0.875 0.896 0.656 -0.827 -0.740 -0.739
## pref_mean_young -0.957 -0.979 -0.862 0.952 0.875 0.971
## pref_mean_middle -0.982 -0.996 -0.840 0.953 0.874 0.962
## pref_mean_old -0.979 -0.989 -0.840 0.951 0.897 0.935
## muslim -0.695 -0.715 -0.588 0.724 0.575 0.791
## net_fiscal 0.779 0.760 0.599 -0.815 -0.627 -0.796
## pref_kurtosis pref_prop_1 pref_prop_4 pref_mean_young
## est_mean 0.883 -0.960 0.856 -0.967
## est_median 0.818 -0.930 0.830 -0.934
## est_sd -0.634 0.189 -0.610 0.354
## est_mad -Inf Inf Inf NaN
## est_skew -0.812 0.942 -0.804 0.936
## est_kurtosis 0.827 -0.460 0.880 -0.618
## est_prop_1 -0.575 0.943 -0.521 0.839
## est_prop_7 0.872 -0.723 0.946 -0.807
## est_mean_young 0.878 -0.952 0.855 -0.969
## est_mean_middle 0.859 -0.970 0.829 -0.965
## est_mean_old 0.902 -0.947 0.875 -0.957
## pref_mean 0.927 -0.946 0.896 -0.979
## pref_median 0.740 -0.790 0.656 -0.862
## pref_sd -0.913 0.929 -0.827 0.952
## pref_mad -0.835 0.794 -0.740 0.875
## pref_skew -0.813 0.978 -0.739 0.971
## pref_kurtosis 1.000 -0.777 0.947 -0.889
## pref_prop_1 -0.777 1.000 -0.717 0.946
## pref_prop_4 0.947 -0.717 1.000 -0.834
## pref_mean_young -0.889 0.946 -0.834 1.000
## pref_mean_middle -0.923 0.945 -0.876 0.977
## pref_mean_old -0.948 0.909 -0.906 0.958
## muslim -0.522 0.835 -0.445 0.731
## net_fiscal 0.612 -0.830 0.547 -0.795
## pref_mean_middle pref_mean_old muslim net_fiscal
## est_mean -0.980 -0.970 -0.722599 0.8064
## est_median -0.948 -0.931 -0.709851 0.7753
## est_sd 0.409 0.469 0.000769 -0.0409
## est_mad -Inf Inf -Inf -Inf
## est_skew 0.931 0.915 0.703297 -0.8038
## est_kurtosis -0.673 -0.731 -0.259524 0.3096
## est_prop_1 0.815 0.759 0.769041 -0.8342
## est_prop_7 -0.843 -0.870 -0.478799 0.5717
## est_mean_young -0.974 -0.961 -0.717043 0.8147
## est_mean_middle -0.974 -0.958 -0.747877 0.8176
## est_mean_old -0.982 -0.979 -0.695430 0.7795
## pref_mean -0.996 -0.989 -0.714579 0.7601
## pref_median -0.840 -0.840 -0.588171 0.5990
## pref_sd 0.953 0.951 0.724072 -0.8148
## pref_mad 0.874 0.897 0.574825 -0.6275
## pref_skew 0.962 0.935 0.790525 -0.7962
## pref_kurtosis -0.923 -0.948 -0.522021 0.6116
## pref_prop_1 0.945 0.909 0.834709 -0.8300
## pref_prop_4 -0.876 -0.906 -0.444788 0.5475
## pref_mean_young 0.977 0.958 0.731489 -0.7954
## pref_mean_middle 1.000 0.989 0.719394 -0.7459
## pref_mean_old 0.989 1.000 0.657093 -0.7103
## muslim 0.719 0.657 1.000000 -0.7295
## net_fiscal -0.746 -0.710 -0.729479 1.0000#relation of SD to Muslim?
GG_scatter(d_reviewer, "muslim", "est_sd", text_pos = "tr")## `geom_smooth()` using formula 'y ~ x'
#fiscal?
GG_scatter(d_reviewer, "net_fiscal", "est_sd", text_pos = "tr")## `geom_smooth()` using formula 'y ~ x'
#regressions
#predict fiscal from estimates
list(
ols(net_fiscal ~ est_mean, data = d_reviewer %>% df_standardize(messages = F)),
ols(net_fiscal ~ est_mean + est_sd, data = d_reviewer %>% df_standardize(messages = F)),
ols(net_fiscal ~ est_prop_1 + est_prop_7, data = d_reviewer %>% df_standardize(messages = F)),
ols(net_fiscal ~ est_mean + est_sd + est_prop_1 + est_prop_7, data = d_reviewer %>% df_standardize(messages = F))
) %>% summarize_models()#predict fiscal from preferences
list(
ols(net_fiscal ~ pref_mean, data = d_reviewer %>% df_standardize(messages = F)),
ols(net_fiscal ~ pref_mean + pref_sd, data = d_reviewer %>% df_standardize(messages = F)),
ols(net_fiscal ~ pref_prop_1 + pref_prop_4, data = d_reviewer %>% df_standardize(messages = F)),
ols(net_fiscal ~ pref_mean + pref_sd + pref_prop_1 + pref_prop_4, data = d_reviewer %>% df_standardize(messages = F))
) %>% summarize_models()#age to predict stereotype accuracy
ols(stereotype_accuracy ~ age + gender + education_num + party_vote, data = d) %>%
ggpredict("age") %>%
plot()
GG_save("figures/age_prediction_model.png")
ols(stereotype_accuracy ~ rcs(age) + gender + education_num + party_vote, data = d) %>%
ggpredict("age") %>%
plot()
GG_save("figures/age_prediction_model_spline.png")
ols(stereotype_accuracy ~ rcs(age), data = d) %>%
ggpredict("age") %>%
plot()
GG_save("figures/age_prediction_model_spline_alone.png")Meta
write_sessioninfo()## R version 3.6.2 (2019-12-12)
## Platform: x86_64-pc-linux-gnu (64-bit)
## Running under: Linux Mint 19.3
##
## Matrix products: default
## BLAS: /usr/lib/x86_64-linux-gnu/blas/libblas.so.3.7.1
## LAPACK: /usr/lib/x86_64-linux-gnu/lapack/liblapack.so.3.7.1
##
## locale:
## [1] LC_CTYPE=en_US.UTF-8 LC_NUMERIC=C
## [3] LC_TIME=en_US.UTF-8 LC_COLLATE=en_US.UTF-8
## [5] LC_MONETARY=en_US.UTF-8 LC_MESSAGES=en_US.UTF-8
## [7] LC_PAPER=en_US.UTF-8 LC_NAME=C
## [9] LC_ADDRESS=C LC_TELEPHONE=C
## [11] LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C
##
## attached base packages:
## [1] stats graphics grDevices utils datasets methods base
##
## other attached packages:
## [1] ggeffects_0.14.1 dkstat_0.08 readxl_1.3.1
## [4] rms_5.1-4 SparseM_1.78 polycor_0.7-10
## [7] lavaan_0.6-5 googlesheets4_0.1.0 kirkegaard_2018.05
## [10] metafor_2.1-0 psych_1.9.12.31 magrittr_1.5
## [13] assertthat_0.2.1 weights_1.0.1 mice_3.8.0
## [16] gdata_2.18.0 Hmisc_4.3-1 Formula_1.2-3
## [19] survival_3.1-11 lattice_0.20-40 forcats_0.5.0
## [22] stringr_1.4.0 dplyr_0.8.5 purrr_0.3.3
## [25] readr_1.3.1 tidyr_1.0.2 tibble_2.1.3
## [28] tidyverse_1.3.0 mediation_4.5.0 sandwich_2.5-1
## [31] mvtnorm_1.1-0 Matrix_1.2-18 MASS_7.3-51.5
## [34] lavaanPlot_0.5.1 ggrepel_0.8.2 ggplot2_3.3.0
## [37] pacman_0.5.1
##
## loaded via a namespace (and not attached):
## [1] backports_1.1.5 plyr_1.8.6 splines_3.6.2
## [4] crosstalk_1.0.0 TH.data_1.0-10 digest_0.6.25
## [7] htmltools_0.4.0 fansi_0.4.1 checkmate_2.0.0
## [10] cluster_2.1.0 modelr_0.1.6 lpSolve_5.6.15
## [13] jpeg_0.1-8.1 colorspace_1.4-1 rvest_0.3.5
## [16] haven_2.2.0 xfun_0.12 crayon_1.3.4
## [19] jsonlite_1.6.1 lme4_1.1-21 zoo_1.8-7
## [22] glue_1.3.2 gtable_0.3.0 MatrixModels_0.4-1
## [25] scales_1.1.0 DBI_1.1.0 Rcpp_1.0.3
## [28] xtable_1.8-4 htmlTable_1.13.3 foreign_0.8-76
## [31] stats4_3.6.2 DT_0.12 htmlwidgets_1.5.1
## [34] httr_1.4.1 DiagrammeR_1.0.5 RColorBrewer_1.1-2
## [37] acepack_1.4.1 ellipsis_0.3.0 pkgconfig_2.0.3
## [40] farver_2.0.3 nnet_7.3-13 dbplyr_1.4.2
## [43] utf8_1.1.4 tidyselect_1.0.0 labeling_0.3
## [46] rlang_0.4.5 later_1.0.0 munsell_0.5.0
## [49] multilevel_2.6 cellranger_1.1.0 tools_3.6.2
## [52] visNetwork_2.0.9 cli_2.0.2 generics_0.0.2
## [55] sjlabelled_1.1.3 broom_0.5.5 evaluate_0.14
## [58] fastmap_1.0.1 yaml_2.2.1 knitr_1.28
## [61] fs_1.3.2 nlme_3.1-145 mime_0.9
## [64] quantreg_5.54 xml2_1.2.5 psychometric_2.2
## [67] compiler_3.6.2 rstudioapi_0.11 png_0.1-7
## [70] reprex_0.3.0 pbivnorm_0.6.0 stringi_1.4.6
## [73] nloptr_1.2.2 vctrs_0.2.4 stringdist_0.9.5.5
## [76] pillar_1.4.3 lifecycle_0.2.0 data.table_1.12.8
## [79] insight_0.8.2 httpuv_1.5.2 R6_2.4.1
## [82] latticeExtra_0.6-29 promises_1.1.0 gridExtra_2.3
## [85] codetools_0.2-16 polspline_1.1.17 boot_1.3-24
## [88] gtools_3.8.1 withr_2.1.2 mnormt_1.5-6
## [91] multcomp_1.4-12 mgcv_1.8-31 parallel_3.6.2
## [94] hms_0.5.3 grid_3.6.2 rpart_4.1-15
## [97] minqa_1.2.4 rmarkdown_2.1 lsr_0.5
## [100] shiny_1.4.0 lubridate_1.7.4 base64enc_0.1-3
## [103] rematch_1.0.1#save the global environment
write_rds(as.list(globalenv()), "data/global_envir.rds")