pairplot: all samples plotted as being 0 on y-axis

Hi, I’ve just run into a situation where pairplot yields this:

while a corresponding pandas.scatter_matrix yields this: The reproducible example below which produces the figures above uses custom data (csv attached: bug.txt) because I guess this behavior somehow has to do with the small values in the example:

import pandas as pd
import seaborn as sns; sns.set(style="ticks", color_codes=True)
data = pd.read_csv('bug.txt') # is csv but needs .txt extension for github
pd.plotting.scatter_matrix(data) # seems to be OK
sns.pairplot(data) # all samples shown as 0 on y-axis

Used versions:

sns.__version__
'0.8.1'
pd.__version__
'0.22.0'
matplotlib.__version__
'2.1.2'

[jbae11]

@MainzerKaiser I came up with this sort of hacky way of doing it, hopefully it helps.

def better_scatter(x, y, **kwargs):
    plt.scatter(x, y, **kwargs)
    margin = (max(y) - min(y)) * 0.1
    plt.ylim(min(y) - margin, max(y) + margin)

g = sns.PairGrid(pp_data)
g = g.map_diag(plt.hist)
g = g.map_offdiag(better_scatter)

[geekoverdose]

Ack. Should anyone else run into this: a quick-fix until the issue is fixed in matplotlib is to just set seaborn’s ylim by hand , e.g. as:

g = sns.pairplot(data)
g.set(ylim=(data.min().min(),data.max().max()))