Plot correct percentages when using plot_counts #17
Conversation
|
Hi! Thank you for identifying this issue and submitting a PR for it. I appreciate your taking the time to improve this library!
I think the reason this behavior exists is because plot_counts is honestly mostly an internal method, and the preferred/main method to use for plotting is plot_likert, which already does the exact conversion you're suggesting: plot-likert/plot_likert/plot_likert.py Lines 222 to 223 in b83b70d That said, since I'll try out your patch when I have a bit more time, but thanks again for submitting it! I'll try to get around to it soon. |
| @@ -52,6 +52,10 @@ def plot_counts( | |||
| # Pad each row/question from the left, so that they're centered around the middle (Neutral) response | |||
| scale_middle = len(scale) // 2 | |||
|
|
|||
| # scales labels to correct percentages | |||
| if plot_percentage: | |||
| counts = likert_percentages(counts, scale, from_count=True) | |||
There was a problem hiding this comment.
One thing I wanted to flag really quickly is that there seems to be some mismatch between the named argument here (from_count) and the one you introduced below (counted).
There was a problem hiding this comment.
One thing I wanted to flag really quickly is that there seems to be some mismatch between the named argument here (from_count) and the one you introduced below (counted).
Yes sorry about that, I changed the name from from_count to counted at the last minute because I thought it sounded a bit cleaner.
I think the reason this behavior exists is because plot_counts is honestly mostly an internal method, and the preferred/main method to use for plotting is plot_likert, which already does the exact conversion you're suggesting
I agree that plot_counts does seem to act as mostly an internal method however, it is a very useful tool to have as an external method and it is mentioned in the documentation as an option (it saved me a lot of time !). Maybe it would make more sense to incorporate plot_counts into plot_likert as an option to be enabled ? Or maybe to encapsulate the behaivour of plot_counts in another function that is exposed ?
Anyway thanks for taking the time to review the pull request - this library is really great !
As pointed out in #17, its behavior is somewhat confusing: > plot_percentages seemed to only just add a percentage sign on to the end of the number of respondents rather than convert the number of responses to percentages As a first step towards fixing that problem, this commit gives the argument a (hopefully) more descriptive name.
In this second part of the fix for #17, a new more intuitive behavior is introduced: given a set of pre-computed counts, setting the `compute_percentages` option will re-calculate the counts as percentages, and display the plot accordingly.
|
I finally got around to fixing this issue in #24 . Thanks again for raising this and proposing a fix! |
Issue
When using plot_count to plot a graph, plot_percentages seemed to only just add a percentage sign on to the end of the number of respondents rather than convert the number of responses to percentages.
Soloution
Modified the likert_percentage function to be able to accept dataframes which are already counted and then added a conditional in plot_count to modify the dataframe accordingly if plot_percentages=True.