0

I'm using seaborn violinplot with inner='box' and with cut=0.

I see the boxplot has a line that extends all the way to the cut. That suggests to me that the boxplot was computed not on the data itself, but on the kde. Is this correct?

nbecker
  • 1,645
  • 5
  • 17
  • 23

1 Answers1

1

No, your interpretation is incorrect. The boxplot is based on the data. When you use cut=0, the KDE extends only to the most extreme datapoints. If you don't have any outliers, the boxplot line and KDE boundary should coincide.

mwaskom
  • 46,693
  • 16
  • 125
  • 127
  • Thanks, but I am confused. In some of my plots the line does not extend to the end of the violinplot. The ends of the lines are supposed to be the extrema of the data, correct? IIUC, the line should always end exactly on the end of the violinplot when cut=0 since both should stop at the extremal points. What am I misunderstanding? – nbecker Aug 04 '16 at 12:55
  • 1
    Please make a reproducible example that demonstrates the problem you are seeing and update the question, otherwise I can't help. – mwaskom Aug 04 '16 at 14:59