We have hard-coded into the code the minimum number of samples that a group must contain to draw a histogram, because histograms look increasingly weird as the number of samples get small (this is true both with and without kernel density estimation). Limit was 200, now set to 60.
Perhaps this should be user-determined. Put a slider under "more settings" to let the user override our threshold.
However, currently, doing that would be complicated, because the back end thinks it knows what the front end will want to plot, and doesn't assign colors to small groups.
It was probably a bad architecture decision in the first place to put the color-selection into the back end anyway. So, although implementing this slider would be extremely easy, a prerequisite should be to reorganize such that the color-selection happens on the front end.
We have hard-coded into the code the minimum number of samples that a group must contain to draw a histogram, because histograms look increasingly weird as the number of samples get small (this is true both with and without kernel density estimation). Limit was 200, now set to 60.
Perhaps this should be user-determined. Put a slider under "more settings" to let the user override our threshold.
However, currently, doing that would be complicated, because the back end thinks it knows what the front end will want to plot, and doesn't assign colors to small groups.
It was probably a bad architecture decision in the first place to put the color-selection into the back end anyway. So, although implementing this slider would be extremely easy, a prerequisite should be to reorganize such that the color-selection happens on the front end.