Seaborn: wrong Y scale with pairgrid

[oneunreadmail]

Following code scales Y axis with small values wrong in half of cases:

import numpy as np
import pandas as pd
import seaborn as sns
import matplotlib.pyplot as plt

df = pd.DataFrame(np.random.randn(100,4), columns=list('ABCD'))
df.A = df.A/1e6
df.B = df.B/1e3
df.D = df.D*1e3

sns.pairplot(df, vars=["A", "B", "C", "D"])

sns.PairGrid(df).map(plt.scatter) will autoscale both axes wrong in every case for small values:

Probably this goes to plt.scatter, because plt.scatter([-1e-6, 1e-6], [-1e-6, 1e-6]) won't autoscale right.

862 is about exactly this, however, I'm still puzzled with half-normal pairplot behavior, seems it should return diag-symmetrical plots even with wrong autoscale in scatter function.

[mwaskom]

I'm still puzzled with half-normal pairplot behavior, seems it should return diag-symmetrical plots even with wrong autoscale in scatter function.

This is probably because the histogram on the diag axes is setting the proper scaling of the x axes?

[oneunreadmail]

This is probably because the histogram on the diag axes is setting the proper scaling of the x axes?

Yeah, you're probably right. Does it happen inside plt as well? This means, it can't (shouldn't) be fixed in seaborn?

[mwaskom]

Yes, I don't understand exactly what the issue with autoscaling is, but this is a problem at the matplotlib layer.

[StevenLi-DS]

I have an issue with distplot and kdeplot in PairGrid with 0.9.0

import seaborn as sns
import pandas as pd
df = sns.load_dataset('tips')
sns.PairGrid(df.dropna())\
   .map_lower(sns.regplot, scatter_kws={'s': 0.5})\
   .map_upper(sns.kdeplot, shade=True)\
   .map_diag(sns.kdeplot)

But using distplot alone will produce

sns.distplot(df.total_bill)

Is there any way to solve this issue?

[thomascarreau]

You want

sns.PairGrid(df.dropna(), diag_sharey=False)

to solve the issue.