Social segregation in Core Discussion Networks

---

Custom functions

• pacage.check: Check if packages are installed (and install if not) in R (source).

fpackage.check <- function(packages) {
    lapply(packages, FUN = function(x) {
        if (!require(x, character.only = TRUE)) {
            install.packages(x, dependencies = TRUE)
            library(x, character.only = TRUE)
        }
    })
}

---

Packages

• tidyverse: if you can’t base them, join them
  
• haven: because we have haven labelled vars in the dataset.
  
• Hmisc: for weighted mean/sd
  
• plotly: for interactive plots

packages = c("tidyverse", "haven", "Hmisc", "plotly")

fpackage.check(packages)

---

Dataset

load(file = "./data/datajt.rda")
load(file = "./data/datajt_nk.rda")

---

Education

datajt %>% 
  filter(leeftijd>24 & !is.na(opl4)) %>%
  group_by(survey_wave, opl4) %>%
  summarise(N = n(),
            cdn_educ_EIM = mean(educ_EI, na.rm=T),
            sd = sd(educ_EI, na.rm=T),
            cdn_educ_EIM_w = wtd.mean(educ_EI, cdn_size, normwt = FALSE, na.rm = TRUE), #perhaps better to set normwt to TRUE
            sd_w = sqrt(wtd.var(educ_EI, cdn_size, normwt = FALSE, na.rm = TRUE))) %>%
  mutate(
    conf.interval = .95,
    se = sd / sqrt(N) ,
    ci = se * qt(conf.interval/2 + .5, N-1),
    se_w = sd_w / sqrt(N) , #is this correct??
    ci_w = se_w * qt(conf.interval/2 + .5, N-1)) %>%   
  na.omit() -> tgc


datajt_nk %>% 
  filter(leeftijd>24) %>%
  group_by(survey_wave, opl4) %>%
  summarise(N = n(),
            cdn_educ_EIM = mean(educ_EI, na.rm=T),
            sd = sd(educ_EI, na.rm=T)) %>%
  mutate(
    se = sd / sqrt(N) ,
    conf.interval = .95,
    ci = se * qt(conf.interval/2 + .5, N-1)) %>% 
  na.omit() -> tgc_nk


tgc$period <- as.numeric(tgc$survey_wave) + 2007
tgc$homogeneity <- round(tgc$cdn_educ_EIM, 2)
tgc$education <- as.factor(tgc$opl4)
levels(tgc$education) <- c("primary + vmbo", "havo/vwo/mbo", "hbo", "university")

tgc_nk$period <- as.numeric(tgc_nk$survey_wave) + 2007
tgc_nk$homogeneity <- round(tgc_nk$cdn_educ_EIM, 2)
tgc_nk$education <- as.factor(tgc_nk$opl4)
levels(tgc_nk$education) <- c("primary + vmbo", "havo/vwo/mbo", "hbo", "university")


datajt %>% 
  filter(leeftijd>24 & !is.na(opl4)) %>%
  group_by(survey_wave, opl4) %>%
  summarise(N = n(),
            cdn_educ_sim = mean(educ_sim, na.rm=T),
            sd = sd(educ_EI, na.rm=T),
            cdn_educ_sim_w = wtd.mean(educ_sim, cdn_size_educ, normwt = FALSE, na.rm = TRUE), #perhaps better to set normwt to TRUE
            sd_w = sqrt(wtd.var(educ_EI, cdn_size_educ, normwt = FALSE, na.rm = TRUE))) %>%
  mutate(
    conf.interval = .95,
    se = sd / sqrt(N) ,
    ci = se * qt(conf.interval/2 + .5, N-1),
    se_w = sd_w / sqrt(N) , #is this correct??
    ci_w = se_w * qt(conf.interval/2 + .5, N-1)) %>%   
  na.omit() -> tgc_op2


datajt_nk %>% 
   filter(leeftijd>24 & !is.na(opl4)) %>%
  group_by(survey_wave, opl4) %>%
  summarise(N = n(),
            cdn_educ_sim = mean(educ_sim, na.rm=T),
            sd = sd(educ_EI, na.rm=T),
            cdn_educ_sim_w = wtd.mean(educ_sim, cdn_size_educ, normwt = FALSE, na.rm = TRUE), #perhaps better to set normwt to TRUE
            sd_w = sqrt(wtd.var(educ_EI, cdn_size_educ, normwt = FALSE, na.rm = TRUE))) %>%
  mutate(
    conf.interval = .95,
    se = sd / sqrt(N) ,
    ci = se * qt(conf.interval/2 + .5, N-1),
    se_w = sd_w / sqrt(N) , #is this correct??
    ci_w = se_w * qt(conf.interval/2 + .5, N-1)) %>%   
  na.omit() -> tgc_nk_op2


tgc$period <- as.numeric(tgc$survey_wave) + 2007
tgc$homogeneity <- round(tgc$cdn_educ_EIM, 2)
tgc$education <- as.factor(tgc$opl4)
levels(tgc$education) <- c("primary + vmbo", "havo/vwo/mbo", "hbo", "university")

tgc_op2$period <- as.numeric(tgc_op2$survey_wave) + 2007
tgc_op2$homogeneity <- round(tgc_op2$cdn_educ_sim, 2)
tgc_op2$education <- as.factor(tgc_op2$opl4)
levels(tgc_op2$education) <- c("primary + vmbo", "havo/vwo/mbo", "hbo", "university")

tgc_nk$period <- as.numeric(tgc_nk$survey_wave) + 2007
tgc_nk$homogeneity <- round(tgc_nk$cdn_educ_EIM, 2)
tgc_nk$education <- as.factor(tgc_nk$opl4)
levels(tgc_nk$education) <- c("primary + vmbo", "havo/vwo/mbo", "hbo", "university")

tgc_nk_op2$period <- as.numeric(tgc_nk_op2$survey_wave) + 2007
tgc_nk_op2$homogeneity <- round(tgc_nk_op2$cdn_educ_sim, 2)
tgc_nk_op2$education <- as.factor(tgc_nk_op2$opl4)
levels(tgc_nk_op2$education) <- c("primary + vmbo", "havo/vwo/mbo", "hbo", "university")

all

plot <- ggplot(tgc, aes(x=period, y=homogeneity, colour=education)) + 
  geom_line(position=position_dodge(0.1)) +
  geom_errorbar(aes(ymin=homogeneity-ci, ymax=homogeneity+ci), width=.6, position=position_dodge(0.1)) +
  geom_point(position=position_dodge(0.1)) + 
  scale_x_continuous(breaks = 2008:2021, limits = c(2007, 2022), labels = as.character(c(2008:2021))) +
  scale_y_continuous(breaks = c(-1, -0.75, -.5, -.25, 0, .25, .5, .75, 1), limits = c(-1,1)) + 
  labs(title = "Educational homogeneity in CDN", caption = "Note: 95% CI \n Source: CentERdata 2021", 
       y = "EI-index", x="period") + 
  scale_colour_hue(name="education ego",    # Legend label, use darker colors
                   #breaks=c("1", "2", "3", "4"),
                   #labels=c("primary + vmbo", "havo/vwo/mbo", "hbo", "university"),
                   l=40)       +              # Use darker colors, lightness=40
  theme(axis.text.x = element_text(angle=45),
        legend.justification=c(1,0),
        legend.position="right")  



#plot
ggplotly(plot) %>%
  layout(margin = list(b=100), annotations = list(x=2021, y = -1.1, showarrow = F, font = list(size=10), text="\n \n  \n  \n \n \n \n \n \n Note: 95% CI \n Source: CentERdata 2021"))

Significance of trend:

• negative trend of ‘primary + vmbo’ and ‘havo/vwo/mbo’
  
• positive trend of ‘university’

datajt %>%
    filter(leeftijd > 24 & !is.na(opl4)) %>%
    with(lm(educ_EI ~ as.factor(opl4) + as.numeric(survey_wave):as.factor(opl4))) %>%
    summary()

#> 
#> Call:
#> lm(formula = educ_EI ~ as.factor(opl4) + as.numeric(survey_wave):as.factor(opl4))
#> 
#> Residuals:
#>     Min      1Q  Median      3Q     Max 
#> -1.3460 -0.3151 -0.1239  0.3933  1.8851 
#> 
#> Coefficients:
#>                                           Estimate Std. Error t value Pr(>|t|)    
#> (Intercept)                              -0.843033   0.009573 -88.059  < 2e-16 ***
#> as.factor(opl4)2                          1.197284   0.013012  92.016  < 2e-16 ***
#> as.factor(opl4)3                          0.654055   0.013693  47.767  < 2e-16 ***
#> as.factor(opl4)4                          0.632496   0.018762  33.712  < 2e-16 ***
#> as.factor(opl4)1:as.numeric(survey_wave) -0.003002   0.001246  -2.409   0.0160 *  
#> as.factor(opl4)2:as.numeric(survey_wave) -0.008217   0.001067  -7.704 1.34e-14 ***
#> as.factor(opl4)3:as.numeric(survey_wave)  0.002036   0.001156   1.762   0.0781 .  
#> as.factor(opl4)4:as.numeric(survey_wave)  0.008592   0.001793   4.791 1.66e-06 ***
#> ---
#> Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
#> 
#> Residual standard error: 0.6061 on 58226 degrees of freedom
#>   (28918 observations deleted due to missingness)
#> Multiple R-squared:  0.3506, Adjusted R-squared:  0.3505 
#> F-statistic:  4490 on 7 and 58226 DF,  p-value: < 2.2e-16

---

no kin

plot <- ggplot(tgc_nk, aes(x=period, y=homogeneity, colour=education)) + 
  geom_line(position=position_dodge(0.1)) +
  geom_errorbar(aes(ymin=homogeneity-ci, ymax=homogeneity+ci), width=.6, position=position_dodge(0.1)) +
  geom_point(position=position_dodge(0.1)) + 
  scale_x_continuous(breaks = 2008:2021, limits = c(2007, 2022), labels = as.character(c(2008:2021))) +
  scale_y_continuous(breaks = c(-1, -0.75, -.5, -.25, 0, .25, .5, .75, 1), limits = c(-1,1)) + 
  labs(title = "Educational homogeneity in CDN", caption = "Note: 95% CI \n Source: CentERdata 2021", 
       y = "EI-index", x="period") + 
  scale_colour_hue(name="education ego",    # Legend label, use darker colors
                   #breaks=c("1", "2", "3", "4"),
                   #labels=c("primary + vmbo", "havo/vwo/mbo", "hbo", "university"),
                   l=40)       +              # Use darker colors, lightness=40
  theme(axis.text.x = element_text(angle=45),
        legend.justification=c(1,0),
        legend.position="right")  


#plot
#plot
ggplotly(plot) %>%
  layout(margin = list(b=100), annotations = list(x=2021, y = -1.1, showarrow = F, font = list(size=10), text="\n \n  \n  \n  \n \n \n \n \n Note: 95% CI \n Source: CentERdata 2021"))

---

all (similarity)

#hist(datajt$educ_sim)


plot <- ggplot(tgc_op2, aes(x=period, y=homogeneity, colour=education)) + 
  geom_line(position=position_dodge(0.1)) +
  geom_errorbar(aes(ymin=homogeneity-ci, ymax=homogeneity+ci), width=.6, position=position_dodge(0.1)) +
  geom_point(position=position_dodge(0.1)) + 
  scale_x_continuous(breaks = 2008:2021, limits = c(2007, 2022), labels = as.character(c(2008:2021))) +
  scale_y_continuous(breaks = c(-1, -0.75, -.5, -.25, 0, .25, .5, .75, 1), limits = c(-1,1)) + 
  labs(title = "Educational homogeneity in CDN", caption = "Note: 95% CI \n Source: CentERdata 2021", 
       y = "S-index", x="period") + 
  scale_colour_hue(name="education ego",    # Legend label, use darker colors
                   #breaks=c("1", "2", "3", "4"),
                   #labels=c("primary + vmbo", "havo/vwo/mbo", "hbo", "university"),
                   l=40)       +              # Use darker colors, lightness=40
  theme(axis.text.x = element_text(angle=45),
        legend.justification=c(1,0),
        legend.position="right")  



#plot
ggplotly(plot) %>%
  layout(margin = list(b=100), annotations = list(x=2021, y = -1.1, showarrow = F, font = list(size=10), text="\n \n  \n \n \n \n \n \n \n Note: 95% CI \n Source: CentERdata 2021"))

---

no kin (similarity)

plot <- ggplot(tgc_nk_op2, aes(x=period, y=homogeneity, colour=education)) + 
  geom_line(position=position_dodge(0.1)) +
  geom_errorbar(aes(ymin=homogeneity-ci, ymax=homogeneity+ci), width=.6, position=position_dodge(0.1)) +
  geom_point(position=position_dodge(0.1)) + 
  scale_x_continuous(breaks = 2008:2021, limits = c(2007, 2022), labels = as.character(c(2008:2021))) +
  scale_y_continuous(breaks = c(-1, -0.75, -.5, -.25, 0, .25, .5, .75, 1), limits = c(-1,1)) + 
  labs(title = "Educational homogeneity in CDN", caption = "Note: 95% CI \n Source: CentERdata 2021", 
       y = "S-index", x="period") + 
  scale_colour_hue(name="education ego",    # Legend label, use darker colors
                   #breaks=c("1", "2", "3", "4"),
                   #labels=c("primary + vmbo", "havo/vwo/mbo", "hbo", "university"),
                   l=40)       +              # Use darker colors, lightness=40
  theme(axis.text.x = element_text(angle=45),
        legend.justification=c(1,0),
        legend.position="right")  



#plot
ggplotly(plot) %>%
  layout(margin = list(b=100), annotations = list(x=2021, y = -1.1, showarrow = F, font = list(size=10), text="\n \n  \n  \n \n \n \n \n \n Note: 95% CI \n Source: CentERdata 2021"))

---

Gender

datajt %>%
    filter(cdn_size > 0 & !is.na(geslacht)) %>%
    group_by(survey_wave, geslacht) %>%
    summarise(N = n(), gender_EI_mean = mean(gender_EI, na.rm = T), sd = sd(gender_EI, na.rm = T)) %>%
    mutate(se = sd/sqrt(N), conf.interval = 0.95, ci = se * qt(conf.interval/2 + 0.5, N - 1)) -> tgc

tgc$period <- as.numeric(tgc$survey_wave) + 2007
tgc$homogeneity <- round(tgc$gender_EI_mean, 2)
tgc$gender <- as.factor(tgc$geslacht)
levels(tgc$gender) <- c("male", "female")

datajt_nk %>%
    filter(cdn_size > 0 & !is.na(geslacht)) %>%
    group_by(survey_wave, geslacht) %>%
    summarise(N = n(), gender_EI_mean = mean(gender_EI, na.rm = T), sd = sd(gender_EI, na.rm = T)) %>%
    mutate(se = sd/sqrt(N), conf.interval = 0.95, ci = se * qt(conf.interval/2 + 0.5, N - 1)) -> tgc_nk

tgc_nk$period <- as.numeric(tgc_nk$survey_wave) + 2007
tgc_nk$homogeneity <- round(tgc_nk$gender_EI_mean, 2)
tgc_nk$gender <- as.factor(tgc_nk$geslacht)
levels(tgc_nk$gender) <- c("male", "female")

all

plot <- ggplot(tgc, aes(x=period, y=homogeneity, colour=gender)) + 
  geom_line(position=position_dodge(0.1)) +
  geom_errorbar(aes(ymin=homogeneity-ci, ymax=homogeneity+ci), width=.6, position=position_dodge(0.1)) +
  geom_point(position=position_dodge(0.1)) + 
  scale_x_continuous(breaks = 2008:2021, limits = c(2007,2022), labels = as.character(c(2008:2021))) +
  scale_y_continuous(breaks = c(-1, -0.75, -.5, -.25, 0, .25, .5, .75, 1), limits = c(-1,1)) + 
  labs(title = "Gender homogeneity in CDN", caption = "Note: 95% CI \n Source: CentERdata 2021", 
       y = "EI-index", x="period") + 
  scale_colour_hue(name="gender ego",    # Legend label, use darker colors
                   #breaks=c("2", "1"),
                   #labels=c("female", "male"),
                   l=40)       +              # Use darker colors, lightness=40
  theme(axis.text.x = element_text(angle=45),
        legend.justification=c(1,0),
        legend.position="right")  

ggplotly(plot) %>%
  layout(margin = list(b=100), annotations = list(x=2021, y = -1, showarrow = F, font = list(size=10), text="\n \n \n \n \n  \n  \n \n \n \n \n Note: 95% CI \n Source: CentERdata 2021"))

---

no kin

plot <- ggplot(tgc_nk, aes(x=period, y=homogeneity, colour=gender)) + 
  geom_line(position=position_dodge(0.1)) +
  geom_errorbar(aes(ymin=homogeneity-ci, ymax=homogeneity+ci), width=.6, position=position_dodge(0.1)) +
  geom_point(position=position_dodge(0.1)) + 
  scale_x_continuous(breaks = 2008:2021, limits = c(2007,2022), labels = as.character(c(2008:2021))) +
  scale_y_continuous(breaks = c(-1, -0.75, -.5, -.25, 0, .25, .5, .75, 1), limits = c(-1,1)) + 
  labs(title = "Gender homogeneity in CDN", caption = "Note: 95% CI \n Source: CentERdata 2021", 
       y = "EI-index", x="period") + 
  scale_colour_hue(name="gender ego",    # Legend label, use darker colors
                   #breaks=c("2", "1"),
                   #labels=c("female", "male"),
                   l=40)       +              # Use darker colors, lightness=40
  theme(axis.text.x = element_text(angle=45),
        legend.justification=c(1,0),
        legend.position="right")  

ggplotly(plot) %>%
  layout(margin = list(b=100), annotations = list(x=2021, y = -1, showarrow = F, font = list(size=10), text="\n \n \n \n  \n \n \n \n \n \n \n Note: 95% CI \n Source: CentERdata 2021"))

---

Age

datajt %>% 
  filter(!is.na(leeftijd) & leeftijd>15) %>%
  group_by(survey_wave, leeftijd_cat13) %>%
  summarise(N = n(),
            cdn_leeftijd_EIM = mean(leeftijd_EI, na.rm=T),
            sd = sd(leeftijd_EI, na.rm=T),
            cdn_leeftijd_EIM_w = wtd.mean(leeftijd_EI, cdn_size_age, normwt = FALSE, na.rm = TRUE), #perhaps better to set normwt to TRUE
            sd_w = sqrt(wtd.var(leeftijd_EI, cdn_size_age, normwt = FALSE, na.rm = TRUE))) %>%
  mutate(
    conf.interval = .95,
    se = sd / sqrt(N) ,
    ci = se * qt(conf.interval/2 + .5, N-1),
    se_w = sd_w / sqrt(N) , #is this correct??
    ci_w = se_w * qt(conf.interval/2 + .5, N-1)) %>%   
  na.omit() -> tgc



datajt_nk %>% 
  filter(!is.na(leeftijd) & leeftijd>15) %>%
  group_by(survey_wave, leeftijd_cat13) %>%
  summarise(N = n(),
            cdn_leeftijd_EIM = mean(leeftijd_EI, na.rm=T),
            sd = sd(leeftijd_EI, na.rm=T),
            cdn_leeftijd_EIM_w = wtd.mean(leeftijd_EI, cdn_size, normwt = FALSE, na.rm = TRUE), #perhaps better to set normwt to TRUE
            sd_w = sqrt(wtd.var(leeftijd_EI, cdn_size, normwt = FALSE, na.rm = TRUE))) %>%
  mutate(
    conf.interval = .95,
    se = sd / sqrt(N) ,
    ci = se * qt(conf.interval/2 + .5, N-1),
    se_w = sd_w / sqrt(N) , #is this correct??
    ci_w = se_w * qt(conf.interval/2 + .5, N-1)) %>%   
  na.omit() -> tgc_nk


test <- cut(as.numeric(datajt$leeftijd),breaks = c(-Inf, 15, 20, 25, 30, 35, 40,45,50,55,60,65,70, Inf))
#levels(test)


tgc$period <- as.numeric(tgc$survey_wave) + 2007
tgc$homogeneity <- round(tgc$cdn_leeftijd_EIM, 2)
tgc$age <- as.factor(tgc$leeftijd_cat13)
levels(tgc$age) <- levels(test)[-1]


tgc_nk$period <- as.numeric(tgc_nk$survey_wave) + 2007
tgc_nk$homogeneity <- round(tgc_nk$cdn_leeftijd_EIM, 2)
tgc_nk$age <- as.factor(tgc_nk$leeftijd_cat13)
levels(tgc_nk$age) <- levels(test)[-1]

datajt %>% 
  filter(!is.na(leeftijd) & leeftijd>15) %>%
  group_by(survey_wave, leeftijd_cat13) %>%
  summarise(N = n(),
            cdn_leeftijd_EIM = mean(leeftijd_sim, na.rm=T),
            sd = sd(leeftijd_sim, na.rm=T),
            cdn_leeftijd_EIM_w = wtd.mean(leeftijd_sim, cdn_size_age, normwt = FALSE, na.rm = TRUE), #perhaps better to set normwt to TRUE
            sd_w = sqrt(wtd.var(leeftijd_sim, cdn_size_age, normwt = FALSE, na.rm = TRUE))) %>%
  mutate(
    conf.interval = .95,
    se = sd / sqrt(N) ,
    ci = se * qt(conf.interval/2 + .5, N-1),
    se_w = sd_w / sqrt(N) , #is this correct??
    ci_w = se_w * qt(conf.interval/2 + .5, N-1)) %>%   
  na.omit() -> tgc_op2



datajt_nk %>% 
  filter(!is.na(leeftijd) & leeftijd>15) %>%
  group_by(survey_wave, leeftijd_cat13) %>%
  summarise(N = n(),
            cdn_leeftijd_EIM = mean(leeftijd_sim, na.rm=T),
            sd = sd(leeftijd_sim, na.rm=T),
            cdn_leeftijd_EIM_w = wtd.mean(leeftijd_sim, cdn_size_age, normwt = FALSE, na.rm = TRUE), #perhaps better to set normwt to TRUE
            sd_w = sqrt(wtd.var(leeftijd_sim, cdn_size_age, normwt = FALSE, na.rm = TRUE))) %>%
  mutate(
    conf.interval = .95,
    se = sd / sqrt(N) ,
    ci = se * qt(conf.interval/2 + .5, N-1),
    se_w = sd_w / sqrt(N) , #is this correct??
    ci_w = se_w * qt(conf.interval/2 + .5, N-1)) %>%   
  na.omit() -> tgc_nk_op2


test <- cut(as.numeric(datajt$leeftijd),breaks = c(-Inf, 15, 20, 25, 30, 35, 40,45,50,55,60,65,70, Inf))
#levels(test)


tgc_op2$period <- as.numeric(tgc_op2$survey_wave) + 2007
tgc_op2$homogeneity <- round(tgc_op2$cdn_leeftijd_EIM, 2)
tgc_op2$age <- as.factor(tgc_op2$leeftijd_cat13)
levels(tgc_op2$age) <- levels(test)[-1]


tgc_nk_op2$period <- as.numeric(tgc_nk_op2$survey_wave) + 2007
tgc_nk_op2$homogeneity <- round(tgc_nk_op2$cdn_leeftijd_EIM, 2)
tgc_nk_op2$age <- as.factor(tgc_nk_op2$leeftijd_cat13)
levels(tgc_nk_op2$age) <- levels(test)[-1]

all

plot <- ggplot(tgc, aes(x=period, y=homogeneity, colour=age)) + 
  geom_line(position=position_dodge(0.1)) +
  geom_errorbar(aes(ymin=homogeneity-ci, ymax=homogeneity+ci), width=.6, position=position_dodge(0.1)) +
  geom_point(position=position_dodge(0.1)) + 
  scale_x_continuous(breaks = 2008:2021, limits = c(2007,2022), labels = as.character(c(2008:2021))) +
  scale_y_continuous(breaks = c(-1, -0.75, -.5, -.25, 0, .25, .5, .75, 1), limits = c(-1,1)) + 
  labs(title = "Age homogeneity", caption = "Note: 95% CI \n Source: CentERdata 2021", 
       y = "EI-index", x="period") + 
  scale_colour_hue(name="age ego",    # Legend label, use darker colors
                   #breaks=c("1", "2"),
                   #labels=levels(test),
                   l=40)       +              # Use darker colors, lightness=40
  theme(axis.text.x = element_text(angle=45),
        legend.justification=c(1,0),
        legend.position="right")  

ggplotly(plot) %>%
  layout(margin = list(b=100), annotations = list(x=2021, y = -1, showarrow = F, font = list(size=10), text="\n \n \n \n  \n \n  \n \n \n \n \n Note: 95% CI \n Source: CentERdata 2021"))

---

no kin

plot <- ggplot(tgc_nk, aes(x=period, y=homogeneity, colour=age)) + 
  geom_line(position=position_dodge(0.1)) +
  geom_errorbar(aes(ymin=homogeneity-ci, ymax=homogeneity+ci), width=.6, position=position_dodge(0.1)) +
  geom_point(position=position_dodge(0.1)) + 
  scale_x_continuous(breaks = 2008:2021, limits = c(2007,2022), labels = as.character(c(2008:2021))) +
  scale_y_continuous(breaks = c(-1, -0.75, -.5, -.25, 0, .25, .5, .75, 1), limits = c(-1,1)) + 
  labs(title = "Age homogeneity", caption = "Note: 95% CI \n Source: CentERdata 2021", 
       y = "EI-index", x="period") + 
  scale_colour_hue(name="age ego",    # Legend label, use darker colors
                   #breaks=c("1", "2"),
                   #labels=levels(test),
                   l=40)       +              # Use darker colors, lightness=40
  theme(axis.text.x = element_text(angle=45),
        legend.justification=c(1,0),
        legend.position="right")  
#plot
ggplotly(plot) %>%
  layout(margin = list(b=100), annotations = list(x=2021, y = -1, showarrow = F, font = list(size=10), text="\n \n \n \n  \n  \n \n \n \n \n \n Note: 95% CI \n Source: CentERdata 2021"))

---

all (similarity)

plot <- ggplot(tgc_op2, aes(x=period, y=homogeneity, colour=age)) + 
  geom_line(position=position_dodge(0.1)) +
  geom_errorbar(aes(ymin=homogeneity-ci, ymax=homogeneity+ci), width=.6, position=position_dodge(0.1)) +
  geom_point(position=position_dodge(0.1)) + 
  scale_x_continuous(breaks = 2008:2021, limits = c(2007,2022), labels = as.character(c(2008:2021))) +
  scale_y_continuous(breaks = c(-1, -0.75, -.5, -.25, 0, .25, .5, .75, 1), limits = c(-1,1))+ 
  labs(title = "Age homogeneity", caption = "Note: 95% CI \n Source: CentERdata 2021", 
       y = "S-index", x="period") + 
  scale_colour_hue(name="age ego",    # Legend label, use darker colors
                   #breaks=c("1", "2"),
                   #labels=levels(test),
                   l=40)       +              # Use darker colors, lightness=40
  theme(axis.text.x = element_text(angle=45),
        legend.justification=c(1,0),
        legend.position="right")  

ggplotly(plot) %>%
  layout(margin = list(b=100), annotations = list(x=2021, y = -1, showarrow = F, font = list(size=10), text="\n \n \n \n  \n  \n \n \n \n \n \n Note: 95% CI \n Source: CentERdata 2021"))

---

no kin (similarity)

plot <- ggplot(tgc_nk_op2, aes(x=period, y=homogeneity, colour=age)) + 
  geom_line(position=position_dodge(0.1)) +
  geom_errorbar(aes(ymin=homogeneity-ci, ymax=homogeneity+ci), width=.6, position=position_dodge(0.1)) +
  geom_point(position=position_dodge(0.1)) + 
  scale_x_continuous(breaks = 2008:2021, limits = c(2007,2022), labels = as.character(c(2008:2021))) +
  scale_y_continuous(breaks = c(-1, -0.75, -.5, -.25, 0, .25, .5, .75, 1), limits = c(-1,1)) + 
  labs(title = "Age homogeneity", caption = "Note: 95% CI \n Source: CentERdata 2021", 
       y = "S-index", x="period") + 
  scale_colour_hue(name="age ego",    # Legend label, use darker colors
                   #breaks=c("1", "2"),
                   #labels=levels(test),
                   l=40)       +              # Use darker colors, lightness=40
  theme(axis.text.x = element_text(angle=45),
        legend.justification=c(1,0),
        legend.position="right")  

ggplotly(plot) %>%
  layout(margin = list(b=100), annotations = list(x=2021, y = -1, showarrow = F, font = list(size=10), text="\n \n \n \n  \n \n  \n \n \n \n \n Note: 95% CI \n Source: CentERdata 2021"))

---

Ethnicity

datajt %>% 
  filter(!is.na(etni3)) %>%
  group_by(survey_wave, etni3) %>%
  summarise(N = n(),
            cdn_etni_EIM = mean(etni_EI, na.rm=T),
            sd = sd(etni_EI, na.rm=T),
            cdn_etni_EIM_w = wtd.mean(etni_EI, cdn_size, normwt = FALSE, na.rm = TRUE), #perhaps better to set normwt to TRUE
            sd_w = sqrt(wtd.var(etni_EI, cdn_size, normwt = FALSE, na.rm = TRUE))) %>%
  mutate(
    conf.interval = .95,
    se = sd / sqrt(N) ,
    ci = se * qt(conf.interval/2 + .5, N-1),
    se_w = sd_w / sqrt(N) , #is this correct??
    ci_w = se_w * qt(conf.interval/2 + .5, N-1)) %>%   
  na.omit() -> tgc

datajt_nk %>% 
  filter(!is.na(etni3)) %>%
  group_by(survey_wave, etni3) %>%
  summarise(N = n(),
            cdn_etni_EIM = mean(etni_EI, na.rm=T),
            sd = sd(etni_EI, na.rm=T),
            cdn_etni_EIM_w = wtd.mean(etni_EI, cdn_size, normwt = FALSE, na.rm = TRUE), #perhaps better to set normwt to TRUE
            sd_w = sqrt(wtd.var(etni_EI, cdn_size, normwt = FALSE, na.rm = TRUE))) %>%
  mutate(
    conf.interval = .95,
    se = sd / sqrt(N) ,
    ci = se * qt(conf.interval/2 + .5, N-1),
    se_w = sd_w / sqrt(N) , #is this correct??
    ci_w = se_w * qt(conf.interval/2 + .5, N-1)) %>%   
  na.omit() -> tgc_nk




tgc$period <- as.numeric(tgc$survey_wave) + 2007
tgc$homogeneity <- round(tgc$cdn_etni_EIM, 2)
tgc$ethnicity <- as.factor(tgc$etni3)
levels(tgc$ethnicity) <- c("Dutch", "Western", "non-Western")

tgc_nk$period <- as.numeric(tgc_nk$survey_wave) + 2007
tgc_nk$homogeneity <- round(tgc_nk$cdn_etni_EIM, 2)
tgc_nk$ethnicity <- as.factor(tgc_nk$etni3)
levels(tgc_nk$ethnicity) <- c("Dutch", "Western", "non-Western")

all

plot <- ggplot(tgc, aes(x=period, y=homogeneity, colour=ethnicity)) + 
  geom_line(position=position_dodge(0.1)) +
  geom_errorbar(aes(ymin=homogeneity-ci, ymax=homogeneity+ci), width=.6, position=position_dodge(0.1)) +
  geom_point(position=position_dodge(0.1)) + 
  scale_x_continuous(breaks = 2011:2021, limits = c(2010,2022), labels = as.character(c(2011:2021))) +
  scale_y_continuous(breaks = c(-1, -0.75, -.5, -.25, 0, .25, .5, .75, 1), limits = c(-1,1)) + 
  labs(title = "Ethnic homogeneity", caption = "Note: 95% CI \n Source: CentERdata 2021", 
       y = "EI-index", x="period") + 
  scale_colour_hue(name="ethnicity ego",    # Legend label, use darker colors
                   #breaks=c("1", "2"),
                   #labels=c("D", "W", "NW"),
                   l=40)       +              # Use darker colors, lightness=40
  theme(axis.text.x = element_text(angle=45),
        legend.justification=c(1,0),
        legend.position="right")  

ggplotly(plot) %>%
  layout(margin = list(b=100), annotations = list(x=2021, y = -1, showarrow = F, font = list(size=10), text="\n \n \n \n  \n  \n \n \n \n \n \n Note: 95% CI \n Source: CentERdata 2021"))

---

no kin

plot <- ggplot(tgc_nk, aes(x=period, y=homogeneity, colour=ethnicity)) + 
  geom_line(position=position_dodge(0.1)) +
  geom_errorbar(aes(ymin=homogeneity-ci, ymax=homogeneity+ci), width=.6, position=position_dodge(0.1)) +
  geom_point(position=position_dodge(0.1)) + 
  scale_x_continuous(breaks = 2011:2021, limits = c(2010,2022), labels = as.character(c(2011:2021))) +
  scale_y_continuous(breaks = c(-1, -0.75, -.5, -.25, 0, .25, .5, .75, 1), limits = c(-1,1)) + 
  labs(title = "Ethnic homogeneity", caption = "Note: 95% CI \n Source: CentERdata 2021", 
       y = "EI-index", x="period") + 
  scale_colour_hue(name="ethnicity ego",    # Legend label, use darker colors
                   #breaks=c("1", "2"),
                   #labels=c("D", "W", "NW"),
                   l=40)       +              # Use darker colors, lightness=40
  theme(axis.text.x = element_text(angle=45),
        legend.justification=c(1,0),
        legend.position="right")  

ggplotly(plot) %>%
  layout(margin = list(b=100), annotations = list(x=2021, y = -1, showarrow = F, font = list(size=10), text="\n \n \n \n  \n \n \n \n \n \n \n Note: 95% CI \n Source: CentERdata 2021"))

---

Take Home Message

• Pronounced segregation in all social dimensions.
  
• Overall, segregation in social networks is quite constant over time
  
• Age segregation is increasing (for the elderly)
  
• Educational segregation is increasing (for the higher educated)

---

---

Copyright © 2022 Jochem Tolsma