Tweaking seaborn.boxplot

Question

I would like to compare a set of distributions of scores (score), grouped by some categories (centrality) and colored by some other (model

Answer 1

It has been a while since this answer has activity, but I'll answer OP's question regarding the weird looking lower-bounds for any people that need help in the future.

Once you set your y-axis to logarithmic scale, it becomes impossible to represent y=0, since log(0) tends to -inf.

Therefore, when the values regarding the lower part of your boxplot are either zero or very close to it the box has that look of seeming to be 'cut in half'.

Needless to say that it's also impossible to represent negative y values in a logarithmic scale.