Grouped bar plot in ggplot with y values based on combination of 2 categorical variables?

Question

I am trying to create a grouped bar plot in ggplot, in which there should be 4 bars per each x value. Here is a subset of my data (actual data is about 4x longer):

Verb_Type,Frame,proportion_type,speaker
mental,V CP,0.209513024,Child
mental,V NP,0.138731597,Child
perception,V CP,0.017167382,Child
perception,V NP,0.387528402,Child
mental,V CP,0.437998087,Parent
mental,V NP,0.144086707,Parent
perception,V CP,0.042695836,Parent
perception,V NP,0.398376853,Parent

What I want is to plot Frame as the x values and proportion_type as the y values, but with the bars based on both Verb_Type and speaker. So for each x value (Frame), there would be 4 bars grouped together - a bar each for the proportion_type value corresponding to mental~child, mental~parent, perception~child, perception~parent. I need for the fill color to be based on Verb_Type, and the fill "texture" (saturation or something) based on speaker. I do not want stacked bars, as it would not accurately represent the data.

I don't want to use facet grids because I find it visually difficult to compare all 4 bars when they're separated into 2 groups. I want to group all the bars together so that the visualization is easier. But I can't figure out how to make the appropriate groupings. Is this something I can do in ggplot, or do I need to manipulate the data before plotting? I tried using melt to reshape the data, but either I was doing it wrong, or that's not what I actually should be doing.

Nate Nate · Accepted Answer · 2017-01-17T03:12:45

I think you are looking for the interaction() (i.e. get all unique pairings) between df$Verb_Type and df$speaker to get the column groupings you are after. You can pass this directly to ggplot or make a new variable ahead of time:

ggplot(df, aes(x = Frame, y = proportion_type,
               group = interaction(Verb_Type, speaker), fill = Verb_Type, alpha = speaker)) +
    geom_bar(stat = "identity", position = "dodge") +
    scale_alpha_manual(values = c(.5, 1))

Or:

df$grouper <- interaction(df$Verb_Type, df$speaker)
ggplot(df, aes(x = Frame, y = proportion_type,
               group = grouper, fill = Verb_Type, alpha = speaker)) +
    geom_bar(stat = "identity", position = "dodge") +
    scale_alpha_manual(values = c(.5, 1))

Grouped bar plot in ggplot with y values based on combination of 2 categorical variables?

1 Answers