2

I can show you better than I can explain:

I use seaborn to draw a jointplot of two time series, but I would like to change the scale of the density so that more of the joint distributions would be visible.

This is what comes out with:

g = sns.jointplot(x1, x2, kind="kde", ratio=5, size=7, space=0)

Pic 1: seaborn jointplot of two time series:

But when I superimpose the observations on top of the jointplot I can see that there are plenty of observations outside the colored areas. I use g.plot_joint to superimpose the scatterplot:

g.plot_joint(plt.scatter, c="k", s=10, linewidth=1, marker="+")

Pic 2: jointplot with scatterplot superimposed

Is there a way to adjust the jointplot in some way to show a larger area of the distribution? I would like to see some coloring even in areas with less dense joint distribution.

Trenton McKinney
  • 56,955
  • 33
  • 144
  • 158
Crebit
  • 367
  • 1
  • 4
  • 14

1 Answers1

5

Just use the n_levels keyword. The default is n_levels=10. It is not well documented, but I found it digging in the source.

g = sns.jointplot(x1, x2, kind="kde", ratio=5, size=7, space=0, n_levels=100)
Rob
  • 3,418
  • 1
  • 19
  • 27
  • Wonderful, thank you Rob! I tried to find something like that from the documents but could not find it. – Crebit Mar 16 '16 at 11:58
  • It still works for me. Of course it only works when `kind='kde'`, since the keyword is passed to the underlying kdeplot. – Rob Dec 14 '18 at 08:37