I have a simple study in which I predicted the likelihood of an outcome as a function of a two-level grouping variable. Now I'd like to display these results in a graph. Initially I was thinking of using a bar plot of the predicted probabilities with confidence intervals, but because bar plots don't show the distribution of the data points, I'm leaning towards other options, like a violin plot. However, I'm not sure if this is possible or even desirable.
-
4$\begingroup$ If your grouping variable has only two levels, then fitted probabilities will only take one of two values, won't they? That will make for a very degenerate violin plot. $\endgroup$– Stephan KolassaJun 9, 2016 at 5:57
-
$\begingroup$ Yes, I wasn't thinking too clearly about this but I realized the same. The other alternative would be to use the violin plot to show the subject-level data instead, showing mean percentage correct on the DV (instead of the trial level data, which is dichotomous) but then the concern is that it's not a pictorial summary representation of my analysis. Also, the violin plot is more about displaying the distributions vs. a clean summary, so it's tricky. $\endgroup$– PanPsychJun 10, 2016 at 2:56
1 Answer
Answered in comments by Stephan Kolassa:
If your grouping variable has only two levels, then fitted probabilities will only take one of two values, won't they? That will make for a very degenerate violin plot.