Open broeder-j opened 2 years ago
There is the correlation plot between two questions. And a bar chart with multiple bars has more information than a diffplot, also differences can be better visually perceived.
a diffplot makes sense for direct comparison (i.e. more/less than). This could be absolute as well as % of that quantity (where results are not % already). This is not the same information content as a barplot.
E.g. imagine you compare: Q1 - A: 95/100 & B 90/100 Q2 - A: 15/100 & B 10/100
now the absolute differences are very similar and will show up as simliar. However in a barplot your eye will be mostly drawn to Q1 having large quantities and Q2 having low quantities overall. Therefore an designated way to look into differences is an absolute good approach (if we have questions/data where this makes sense)
Currently we do not have two questions in the same bar plot. but only one question with different answers and research fields. If we are talking about differences between two Questions, I do not know how to do that, and what it should mean.
The counterargument is: the plot is interactive, zoom in, then it looks the same. Or hover over it, then you see the numbers. If we display always some relative numbers then the scale is always comparable for all questions. The diff plot will loose the information that the absolute numbers are so different and a diffplot only works if there is only an A&B and not C&D.
Of course one could plot everything as a difference/relative to a certain reference, but then the question is what to choose as the reference and what meaning that should have.
Currently we do not have two questions in the same bar plot. but only one question with different answers and research fields. If we are talking about differences between two Questions, I do not know how to do that, and what it should mean.
not questions but answers.
The diff plot will loose the information that the absolute numbers are so different and a diffplot only works if there is only an A&B and not C&D
no information is lost - its just another way of offering to look at the data. It is e.g. specifically interesting when comparing adjacent research fields.
Add a button (or some other widgets) which adds a Diffplot, if only two Datasets are shown.
The diffplot is a barchart displaying the difference of the two datasets.