Question

I am creating stacked bar charts on subsets of data using a loop. I’m exporting one stacked bar chart for each variable for each school. Its all working successfully – each chart shows only the subset of data – except that the x axis continues to be labeled for every school. So I have a big long chart area with blanks, and a stacked bar chart over the school for which the data is subsetted.

Here is sample of my data:

label    variable    class    percent
SchoolA   Optimism    High     67
SchoolA   Optimism    Med      33
SchoolA   Optimism    Low      20
SchoolA   SelfEsteem  High     84
SchoolA   SelfEsteem  Med      12
SchoolA   SelfEsteem  Low      4
SchoolB   Optimism    High     60
SchoolB   Optimism    Med      21
SchoolB   Optimism    Low      19
SchoolB   SelfEsteem  High     20
SchoolB   SelfEsteem  Med      42
SchoolB   SelfEsteem  Low      38

…which carries on for several hundred more variables and schools.

In general I’m doing this:

Create an array of values for each school and variable

schools<-unique(df1$label)  
variables<-unique(df1$variable)

Function that plots each subset of data as a stacked bar chart

doPlot<-function(subdf){  
    ggplot(subdf,aes(x=label,y=percent,fill=factor(class)))+  
    geom_bar(stat="identity")}

Runs doPlot using each subset of data

for(i in 1:length(schools)){  
    for (j in 1:length(variables)){  
        subdf<-data.frame(subset(df1,label==schools[i]&variable==variables[j]))  
        doPlot(subdf)}}

Why is the chart creating and labeling an x-axis for the original data set and not only for the subset of data? I am a R newbie, but I’ve been searching and trying things out for a while and I’m stumped.

Answer 1

It’s because when you subset a dataframe with a factor column (df1$label is a factor with levels ‘SchoolA’ and ‘SchoolB’), the label column still has multiple levels, even though there is currently only one unique label in it.

e.g.:

df2 <- subset(df1,label=='SchoolA'&variable=='Optimism')
df2$label
# [1] SchoolA SchoolA SchoolA
# Levels: SchoolA SchoolB

See how even though df2$label is ‘SchoolA’ only, there are still two levels? ggplot uses the levels to draw the graphs.

?subset mentions :

Factors may have empty levels after subsetting; unused levels are
not automatically removed. See ‘droplevels’ for a way to drop all
unused levels from a data frame.

Then going to ?droplevels: “used to drop unused levels from .. factors in a data frame”.

So try doing:

subdf <- droplevels(data.frame(subset(df1,label==schools[i]&variable==variables[j])))

Then you’ll get just one variable on the x axis, the current school.

(As an aside – is a stacked bar chart the best way to represent these? In particular, I notice that SchoolA’s ‘Optimism’ adds up to 120% overall, and the stacked bar chart seems to enforce this understanding of the data. If that’s not how you want the data to be interpreted, perhaps you could consider a different way of presenting it?)