Alexey Bessudnov 27 March 2020
Longitudinal data: political interest
Example from class 5.
library(tidyverse)
# Read data and select variables
Ind1 <- read_tsv("data/UKDA-6614-tab/tab/ukhls_w1/a_indresp.tab")
Ind1 <- Ind1 %>%
select(pidp, a_sex_dv, a_age_dv, a_vote6)
Ind2 <- read_tsv("data/UKDA-6614-tab/tab/ukhls_w2/b_indresp.tab")
Ind2 <- Ind2 %>%
select(pidp, b_sex_dv, b_age_dv, b_vote6)
Ind3 <- read_tsv("data/UKDA-6614-tab/tab/ukhls_w3/c_indresp.tab")
Ind3 <- Ind3 %>%
select(pidp, c_sex_dv, c_age_dv, c_vote6)
Wide <- Ind1 %>%
full_join(Ind2, by = "pidp") %>%
full_join(Ind3, by = "pidp")
Wide## # A tibble: 70,586 x 10
## pidp a_sex_dv a_age_dv a_vote6 b_sex_dv b_age_dv b_vote6 c_sex_dv c_age_dv
## <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
## 1 6.80e7 1 39 3 NA NA NA NA NA
## 2 6.80e7 1 59 2 1 60 2 1 61
## 3 6.80e7 2 39 4 2 40 4 2 41
## 4 6.80e7 2 17 4 NA NA NA 2 19
## 5 6.80e7 2 72 4 2 73 4 2 74
## 6 6.80e7 2 57 1 2 58 2 2 59
## 7 6.80e7 1 59 -7 NA NA NA NA NA
## 8 6.80e7 1 25 -7 NA NA NA NA NA
## 9 6.80e7 1 22 -7 NA NA NA NA NA
## 10 6.80e7 2 38 3 2 39 4 NA NA
## # … with 70,576 more rows, and 1 more variable: c_vote6 <dbl>
# convert to long
Long <- Wide %>%
gather(a_sex_dv:c_vote6, key = "variable", value = "value") %>%
separate(variable, into = c("wave", "variable"), sep = "_", extra = "merge") %>%
spread(key = variable, value = value)
Long## # A tibble: 211,758 x 5
## pidp wave age_dv sex_dv vote6
## <dbl> <chr> <dbl> <dbl> <dbl>
## 1 280165 a NA NA NA
## 2 280165 b 31 2 2
## 3 280165 c 32 2 -7
## 4 541285 a NA NA NA
## 5 541285 b NA NA NA
## 6 541285 c 25 1 -7
## 7 541965 a NA NA NA
## 8 541965 b NA NA NA
## 9 541965 c 23 2 -7
## 10 665045 a NA NA NA
## # … with 211,748 more rows
Long %>%
count(vote6)## # A tibble: 9 x 2
## vote6 n
## <dbl> <int>
## 1 -9 263
## 2 -7 10973
## 3 -2 130
## 4 -1 134
## 5 1 14178
## 6 2 48427
## 7 3 41235
## 8 4 39915
## 9 NA 56503
You may want to use some inequality/heterogeneity measure to explore the stability of political interest at individual level (variance? Theil’s enthropy?). You may also want to visualise changes in political interest.
If you treat political interest as categorical it is possible to use sequence anaysis: http://traminer.unige.ch (this would be a complex analysis and you do not have to do this).
To determine how political interest changes depending on the electoral cycle you’d need to identify the interview dates.
Ind1 <- read_tsv("data/UKDA-6614-tab/tab/ukhls_w1/a_indresp.tab")
Ind1 %>%
select(pidp, a_intdatd_dv, a_intdatm_dv, a_intdaty_dv) %>%
print(n = 10)## # A tibble: 50,994 x 4
## pidp a_intdatd_dv a_intdatm_dv a_intdaty_dv
## <dbl> <dbl> <dbl> <dbl>
## 1 68001367 19 1 2009
## 2 68004087 8 1 2009
## 3 68006127 9 1 2009
## 4 68006135 9 1 2009
## 5 68006807 16 1 2009
## 6 68007487 22 1 2009
## 7 68007491 22 1 2009
## 8 68007495 22 1 2009
## 9 68007499 22 1 2009
## 10 68008167 23 1 2009
## # … with 5.098e+04 more rows
Note that the timing of different waves can overlap.
Childbirth and satisfaction
The key thing here is to identify new childbirths. Note the newborn data sets: https://www.understandingsociety.ac.uk/documentation/mainstage/dataset-documentation/search/datafile/newborn. You want to create a data frame that includes only new mothers and fathers.
newborn2 <- read_tsv("data/UKDA-6614-tab/tab/ukhls_w2/b_newborn.tab")At this stage you may want to filter out those who don’t live with their children or whose children died.
# identify new parents in wave 2
newparent2 <- Ind2 %>%
semi_join(newborn2, by = "pidp")
newparent2 %>%
ggplot(aes(x = b_age_dv)) +
geom_density() +
facet_wrap(~ b_sex_dv)
We can also look at how political interest changes in the wave following the childbirth.
newparent2 <- newparent2 %>%
left_join(Ind3, by = "pidp") %>%
filter(c_vote6 > 0 & b_vote6 > 0) %>%
mutate(polIntChange = c_vote6 - b_vote6)
qplot(newparent2$polIntChange)
newparent2 %>%
group_by(b_sex_dv) %>%
summarise(
meanChange = mean(polIntChange, na.rm = TRUE)
)## # A tibble: 2 x 2
## b_sex_dv meanChange
## <dbl> <dbl>
## 1 1 0.0515
## 2 2 0.0690
Things will become more complex when you’ve got the data from all waves.
Combination of individual- and household-level data: household composition by ethnic group.
Egoalt1 <- read_tsv("data/UKDA-6614-tab/tab/ukhls_w1/a_egoalt.tab")Let us look at the household with a_hidp 68006123.
oneHH <- Egoalt1 %>%
filter(a_hidp == 68006123) %>%
select(a_hidp, pidp, apidp, a_relationship_dv, a_sex, a_asex)
oneHH## # A tibble: 12 x 6
## a_hidp pidp apidp a_relationship_dv a_sex a_asex
## <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
## 1 68006123 68006127 68006131 2 2 1
## 2 68006123 68006127 68006135 9 2 2
## 3 68006123 68006127 68006139 9 2 2
## 4 68006123 68006131 68006127 2 1 2
## 5 68006123 68006131 68006135 12 1 2
## 6 68006123 68006131 68006139 12 1 2
## 7 68006123 68006135 68006127 4 2 2
## 8 68006123 68006135 68006131 7 2 1
## 9 68006123 68006135 68006139 14 2 2
## 10 68006123 68006139 68006127 4 2 2
## 11 68006123 68006139 68006131 7 2 1
## 12 68006123 68006139 68006135 14 2 2
This is a household with four people: a couple with two children (so the relationships are 2 (partner/cohabitee), 9 (parent), 12 (step-parent), 4 (natural son/daughter), 7 (stepson/stepdaughter), 14 (natural brother/sister). More specifically, it’s a woman with two daughters and their step-father, the woman’s partner (they are not married).
The key thing here is to write a code that identifies a household type from this information. Then you can look at the distribution of the types by ethnic group. Note that ethnic group is an individual, not household characteristics, so you will need to develop some rules on how you deal with ethnically mixed households.
As an example, let us calculate the number of people in the household by ethnic group.
Ind1new <- read_tsv("data/UKDA-6614-tab/tab/ukhls_w1/a_indresp.tab")
Ind1new <- Ind1new %>%
select(pidp, a_hidp, a_sex_dv, a_age_dv, a_racel_dv)
Ind1new %>%
group_by(a_hidp) %>%
mutate(nhh = n()) %>%
group_by(a_racel_dv) %>%
summarise(hhsize = mean(nhh, na.rm = TRUE))## # A tibble: 19 x 2
## a_racel_dv hhsize
## <dbl> <dbl>
## 1 -9 2.72
## 2 1 1.99
## 3 2 1.89
## 4 3 1.56
## 5 4 1.97
## 6 5 1.96
## 7 6 2.08
## 8 7 2.18
## 9 8 2.16
## 10 9 2.55
## 11 10 2.73
## 12 11 2.87
## 13 12 2.06
## 14 13 2.20
## 15 14 1.77
## 16 15 2.04
## 17 16 1.99
## 18 17 2.19
## 19 97 2.01
This is actually not quite correct (since ethnicity is an individual-level variable) and you will be able to do better than this.
Let us get some more details on household 68006123.
oneHH %>%
left_join(Ind1new, by = c("a_hidp", "pidp"))## # A tibble: 12 x 9
## a_hidp pidp apidp a_relationship_… a_sex a_asex a_sex_dv a_age_dv
## <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
## 1 6.80e7 6.80e7 6.80e7 2 2 1 2 39
## 2 6.80e7 6.80e7 6.80e7 9 2 2 2 39
## 3 6.80e7 6.80e7 6.80e7 9 2 2 2 39
## 4 6.80e7 6.80e7 6.80e7 2 1 2 NA NA
## 5 6.80e7 6.80e7 6.80e7 12 1 2 NA NA
## 6 6.80e7 6.80e7 6.80e7 12 1 2 NA NA
## 7 6.80e7 6.80e7 6.80e7 4 2 2 2 17
## 8 6.80e7 6.80e7 6.80e7 7 2 1 2 17
## 9 6.80e7 6.80e7 6.80e7 14 2 2 2 17
## 10 6.80e7 6.80e7 6.80e7 4 2 2 NA NA
## 11 6.80e7 6.80e7 6.80e7 7 2 1 NA NA
## 12 6.80e7 6.80e7 6.80e7 14 2 2 NA NA
## # … with 1 more variable: a_racel_dv <dbl>
So the woman is 39 years old and of her daughters is 17. They are both White British. The male partner and the other daughter were not interviewed. We might get some more informatiom about them from the indall file.
Indall1 <- read_tsv("data/UKDA-6614-tab/tab/ukhls_w1/a_indall.tab") %>%
filter(a_hidp == 68006123) %>%
select(a_hidp, pidp, a_sex_dv, a_age_dv, a_ethn_dv)
Indall1## # A tibble: 4 x 5
## a_hidp pidp a_sex_dv a_age_dv a_ethn_dv
## <dbl> <dbl> <dbl> <dbl> <dbl>
## 1 68006123 68006127 2 39 1
## 2 68006123 68006131 1 34 1
## 3 68006123 68006135 2 17 1
## 4 68006123 68006139 2 14 1
The male partner is 34 and the other daughter is 14. They are also White British.
You want to use the indall rather than indresp files for this question.
You may also want to use household composition variables from hhresp files: https://www.understandingsociety.ac.uk/documentation/mainstage/dataset-documentation/search/datafile/hhresp
Note that some household only have one person so they won’t be reflected in the egoalt files.