ggplot2 problems with using ...prop... and grouping bar graph by another category

Question

StudentData <- data.frame(gender = sample( c("male","female"), 100, replace=TRUE),
              degree = sample( c("Associates", "Masters", "PhD"), 100, replace=TRUE),
              category = sample( c("Audit", "Credit"), 100, replace=TRUE))

In the following dataset, I am trying to create a bar graph that plots the percentage of the sample which have an Associate's, Master's, or PhD, separated by gender (done by using facet_grid() ). This is what I have generated so far:

StudentData %>% ggplot(., aes(x=degree, group=gender)) + 
            geom_bar(aes(y=..prop..), stat="count", position=position_dodge()) +
            geom_text(aes(label=scales::percent(round(..prop..,2)), 
            y=..prop..), stat="count", vjust=-.5) +
            scale_y_continuous(limits=c(0,1),labels = scales::percent) +
            ylab("Percent of Sample") +
            facet_grid(~gender)

However, I would also like to display the difference between the groups "Audit" and "Credit" on each graph as side by bars. Yet, when I add "fill=category" to the aesthetics of ggplot, nothing changes:

StudentData %>% ggplot(., aes(x=degree, group=gender, fill=category)) + 
            geom_bar(aes(y=..prop..), stat="count", position=position_dodge()) +
            geom_text(aes(label=scales::percent(round(..prop..,2)), 
            y=..prop..), stat="count", vjust=-.5) +
            scale_y_continuous(limits=c(0,1),labels = scales::percent) +
            ylab("Percent of Sample") +
            facet_grid(~gender)

I realize that usually this is accomplished using geom_bar(stat="identity", position=position_dodge()) but when I change stat="identity", the following error message appears:

Error in FUN(X[[i]], ...) : object 'prop' not found

Any idea how to have a facet graph, use special characters such as ..prop.. AND add another fill to a ggplot2 graph?

I think you'll have better luck if you do your data manipulation in `dplyr` and your plotting in `ggplot2`, rather than trying to lean on the simple built-in data manipulation functions in `ggplot2` to do complex tasks. — Gregor Thomas, Jan 03 '18 at 23:28
Gregor, could you explain which part you are suggesting that I use dplyr for? — Syrah.Sharpe, Jan 04 '18 at 00:08
Calculating the proportions at whichever grouping levels you want. — Gregor Thomas, Jan 04 '18 at 00:10

score 3 · Accepted Answer · answered Jan 03 '18 at 23:46

3

You could create the necessary four groups (instead of two) like this:

StudentData %>% 
  ggplot(., aes(x=degree, group=interaction(gender, category), fill=category)) + 
  geom_bar(aes(y=..prop..), stat="count", position=position_dodge()) +
  geom_text(aes(label=scales::percent(round(..prop..,2)), 
                y=..prop..), stat="count", vjust=-.5, position=position_dodge(.9)) +
  scale_y_continuous(limits=c(0,1),labels = scales::percent) +
  ylab("Percent of Sample") +
  facet_grid(~gender)

answered Jan 03 '18 at 23:46

lukeA

53,097
5
97
100

However, the proportions do not make sense... how are they calculated? – Syrah.Sharpe Jan 04 '18 at 21:28
@Syrah.Sharpe Obviously, each group sums up to 100%. – lukeA Jan 04 '18 at 22:07
@lukeA I think he meant to have a chart were the sum of all the levels is 100%. `interaction()` is redundant anyway, since `group=category` is enough to produce that output – Dambo Sep 10 '18 at 23:17
2

With ggplot2 version 3.x.x you use `stat(prop)` in place of `..prop..` – Jeffrey Girard Jan 22 '19 at 20:59

ggplot2 problems with using ...prop... and grouping bar graph by another category

1 Answers1

Linked