Modifying Visualizations of Amounts Example
Bar Plot Basics
Let’s start with a basic bar plot for Religion
First, let’s look at the count/frequency statistics for the religion variable.
gss %>%
group_by(relig) %>%
summarise(count = n())# A tibble: 15 x 2
relig count
<chr> <int>
1 Buddhism 147
2 Catholic 5124
3 Christian 689
4 Don't know 15
5 Hinduism 71
6 Inter-nondenominational 109
7 Jewish 388
8 Moslem/islam 104
9 Native american 23
10 No answer 93
11 None 3523
12 Orthodox-christian 95
13 Other 224
14 Other eastern 32
15 Protestant 10846Here is the default graphic for a barplot of the counts. Notice, if we use geom_bar() it will automatically summarize for us.
gss %>%
ggplot(aes(x=relig)) +
geom_bar()
Notice that going with just the default is not going to give us a great graph. We need to consider some modifications.
Using forcats
The forcats package allows us to work directly with categorical variables in ways that make applying Gestalt Principle.
Learn more here: https://forcats.tidyverse.org/
Reordering Variable
We can use fct_infreq to apply the principle of simplicity for comparison.
gss %>%
ggplot(aes(x=fct_infreq(relig))) +
geom_bar()
We can use fct_rev to do the reverse order.
gss %>%
ggplot(aes(x=fct_rev(fct_infreq(relig)))) +
geom_bar()
Collapsing Levels
We can use fct_collapse to redefine our factor levels. Remember - doing this is a form of analysis because you are making some assumptions about what groups go together!
gss %>%
mutate(religion = fct_collapse(relig,
Christian = c("Catholic",
"Christian",
"Orthodox-christian",
"Protestant"),
Eastern = c("Buddhism",
"Hinduism",
"Other eastern"),
Jewish = c("Jewish"),
Muslim = c("Moslem/islam"),
None = c("None"),
Unknown = c("No answer",
"Don't know"),
Other = c("Other",
"Native american",
"Inter-nondenominational"))) %>%
ggplot(aes(x=fct_infreq(religion))) +
geom_bar() 
Creating Groups
Use facet_wrap to create groups of visualizations based on a specific variable.
gss %>%
mutate(religion = fct_collapse(relig,
Christian = c("Catholic",
"Christian",
"Orthodox-christian",
"Protestant"),
Eastern = c("Buddhism",
"Hinduism",
"Other eastern"),
Jewish = c("Jewish"),
Muslim = c("Moslem/islam"),
None = c("None"),
Unknown = c("No answer",
"Don't know"),
Other = c("Other",
"Native american",
"Inter-nondenominational"))) %>%
ggplot(aes(x=fct_infreq(religion))) +
geom_bar() +
facet_wrap(~marital, nrow=2)
Do we need to do a bit of data cleaning? Should we consider something other than count to display here? Should we switch the variable?
gss %>%
mutate(religion = fct_collapse(relig,
Christian = c("Catholic",
"Christian",
"Orthodox-christian",
"Protestant"),
Eastern = c("Buddhism",
"Hinduism",
"Other eastern"),
Jewish = c("Jewish"),
Muslim = c("Moslem/islam"),
None = c("None"),
Unknown = c("No answer",
"Don't know"),
Other = c("Other",
"Native american",
"Inter-nondenominational"))) %>%
group_by(religion, marital) %>%
summarise(n=n()) %>%
group_by(religion) %>%
mutate(proportion = n/sum(n)) %>%
filter(marital != "No answer") %>%
ggplot(aes(x=fct_reorder(marital, proportion), y=proportion, fill = marital)) +
geom_col() +
coord_flip() +
facet_wrap(~religion, nrow=3) +
guides(fill = FALSE)`summarise()` has grouped output by 'religion'. You can override using the
`.groups` argument.Warning: `guides(<scale> = FALSE)` is deprecated. Please use `guides(<scale> =
"none")` instead.
Political Affiliation by Religion Group
gss %>%
mutate(religion = fct_collapse(relig,
Christian = c("Catholic",
"Christian",
"Orthodox-christian",
"Protestant"),
Eastern = c("Buddhism",
"Hinduism",
"Other eastern"),
Jewish = c("Jewish"),
Muslim = c("Moslem/islam"),
None = c("None"),
Unknown = c("No answer",
"Don't know"),
Other = c("Other",
"Native american",
"Inter-nondenominational"))) %>%
mutate(part_aff = fct_collapse(partyid,
Conservative = c("Strong republican",
"Not str republican",
"Ind,near rep"),
Liberal = c("Strong democrat",
"Not str democrat",
"Ind,near dem")) %>%
fct_lump()) %>%
group_by(religion, part_aff) %>%
summarise(n=n()) %>%
group_by(religion) %>%
mutate(proportion = n/sum(n)) %>%
mutate(religion = factor(religion, level = c("Jewish", "Muslim", "Eastern",
"None","Other",
"Christian", "Unknown"))) %>%
ggplot(aes(x= religion, y = proportion, fill = part_aff)) +
geom_col(position = position_dodge()) + #position = position_stack()
labs(x = "Religion Groups", y = "Proportion within Religious Group", fill = "Party Affiliation") +
scale_fill_viridis_d() +
theme_bw() +
theme(legend.position = "top")`summarise()` has grouped output by 'religion'. You can override using the
`.groups` argument.
gss %>%
mutate(religion = fct_collapse(relig,
Christian = c("Catholic",
"Christian",
"Orthodox-christian",
"Protestant"),
Eastern = c("Buddhism",
"Hinduism",
"Other eastern"),
Jewish = c("Jewish"),
Muslim = c("Moslem/islam"),
None = c("None"),
Unknown = c("No answer",
"Don't know"),
Other = c("Other",
"Native american",
"Inter-nondenominational"))) %>%
mutate(part_aff = fct_collapse(partyid,
Conservative = c("Strong republican",
"Not str republican",
"Ind,near rep"),
Liberal = c("Strong democrat",
"Not str democrat",
"Ind,near dem")) %>%
fct_lump()) %>%
filter(religion == "None") %>%
#group_by(year, part_aff) %>%
#summarise(n=n()) %>%
#group_by(year) %>%
#mutate(proportion = n/sum(n)) %>%
ggplot(aes(x= year, fill = part_aff)) +
geom_bar() + #position = position_dodge()
labs(x = "Year", y = "Proportion within Christian Group", fill = "Party Affiliation") +
scale_fill_viridis_d() +
theme_bw() +
theme(legend.position = "top")
Average Age by Religion
gss %>%
mutate(religion = fct_collapse(relig,
Christian = c("Catholic",
"Christian",
"Orthodox-christian",
"Protestant"),
Eastern = c("Buddhism",
"Hinduism",
"Other eastern"),
Jewish = c("Jewish"),
Muslim = c("Moslem/islam"),
None = c("None"),
Unknown = c("No answer",
"Don't know"),
Other = c("Other",
"Native american",
"Inter-nondenominational"))) %>%
group_by(religion) %>%
summarise(Avg_age = mean(age, na.rm=TRUE)) %>%
ggplot(aes(x=fct_reorder(religion, Avg_age), y=Avg_age)) +
geom_col(aes(fill=religion)) +
coord_flip() +
guides(fill = FALSE) +
scale_fill_viridis_d() +
labs(x = "Religion Groups", y = "Average Age (years)") +
theme_bw()Warning: `guides(<scale> = FALSE)` is deprecated. Please use `guides(<scale> =
"none")` instead.
library(see)Warning: package 'see' was built under R version 4.1.3gss %>%
mutate(religion = fct_collapse(relig,
Christian = c("Catholic",
"Christian",
"Orthodox-christian",
"Protestant"),
Eastern = c("Buddhism",
"Hinduism",
"Other eastern"),
Jewish = c("Jewish"),
Muslim = c("Moslem/islam"),
None = c("None"),
Unknown = c("No answer",
"Don't know"),
Other = c("Other",
"Native american",
"Inter-nondenominational"))) %>%
drop_na(age) %>%
ggplot(aes(x = fct_reorder(religion, age), y = age)) +
geom_boxplot(aes(fill = religion), alpha = 0.5) +
geom_violinhalf(aes(fill = religion), alpha = 0.2) + #from package "see"
labs(x = "Religion Groups", y = "Age (years)") +
coord_flip() +
scale_fill_viridis_d() +
guides(fill = FALSE, color=FALSE) +
theme_bw()Warning: `guides(<scale> = FALSE)` is deprecated. Please use `guides(<scale> =
"none")` instead.
Religion by Income Level
gss %>%
mutate(religion = fct_collapse(relig,
Christian = c("Catholic",
"Christian",
"Orthodox-christian",
"Protestant"),
Eastern = c("Buddhism",
"Hinduism",
"Other eastern"),
Jewish = c("Jewish"),
Muslim = c("Moslem/islam"),
None = c("None"),
Unknown = c("No answer",
"Don't know"),
Other = c("Other",
"Native american",
"Inter-nondenominational"))) %>%
filter(!rincome %in% c("Refused","Not applicable","No answer","Don't know")) %>%
mutate(rincome = factor(rincome, levels = c("Lt $1000", "$1000 to 2999",
"$3000 to 3999", "$4000 to 4999",
"$5000 to 5999", "$6000 to 6999",
"$7000 to 7999", "$8000 to 9999",
"$10000 - 14999", "$15000 - 19999",
"$20000 - 24999", "$25000 or more"))) %>%
mutate(rincome = fct_collapse(rincome, '$0 to 4999' = c("Lt $1000", "$1000 to 2999",
"$3000 to 3999", "$4000 to 4999"),
'$5000 to 9999' = c("$5000 to 5999", "$6000 to 6999",
"$7000 to 7999", "$8000 to 9999"))) %>%
group_by(religion, rincome) %>%
summarise(n=n()) %>%
group_by(religion) %>%
mutate(proportion = n/sum(n)) %>%
ggplot(aes(x=religion , y=rincome, fill = proportion)) +
geom_tile()+
scale_fill_viridis_c(option = "plasma") +
labs(x = "Religion Groups", y = "Income Level", fill="Proportion") +
theme_bw()`summarise()` has grouped output by 'religion'. You can override using the
`.groups` argument.