quaquel / EMAworkbench

plot_pair_wise_scatter only works properly with categorical parameters if the data has been imported with load_results, if the data was generated directly with perform_experiments, you run into some dtype issues. This is because there is a type conversion in load_results

EMAworkbench/ema_workbench/util/utilities.py

Line 68 in 5628cc1

experiments[name] = experiments[name].astype("category")

that is expected/necessary for correctly drawing boxes in plot_pair_wise_scatter.

In a normal EMA workflow, this probably isn't an issue, since normally you would save/load data before analysing it, but when debugging improved plotting functions, it can be a problem :)

Raising this issue for awareness. I know in the TODO there is something mentioned about moving from the whole category business to Enum or NamedTuple, so maybe this can be taken along there.

This was previously discussed in #72.

Would be good to address this as part of #232

Pairs plotting dependent on load_results