regplot and xscale('log') don't get along without xlim

Hi, When running:

sns.regplot('r', 'in_final_rate', data, ci=None, x_ci=None, logistic=True)
plt.xscale('log')

I get a malformed x-axis. However, if I add xlim():

sns.regplot('r', 'in_final_rate', data, ci=None, x_ci=None, logistic=True)
plt.xlim(1e-4, 1e-1)
plt.xscale('log')

Then it looks fine.

See notebook in this gist, which also includes a data file.

[zherebetskyy]

The log-plot usually looks better if

• the borders are set starting from some value (say 1)
• the values of arguments have to be positive

Something like this should work

import matplotlib.pyplot as plt 
import seaborn as sns
x_column = 'col1' 
y_column = 'col2'
fg = sns.FacetGrid(data=pdf_clean[(pdf_clean['col1']>1) & (pdf_clean['col2']>1)], hue='col3', size=5, aspect=1.41, palette='muted')
fg.map(plt.scatter, x_column, y_column, alpha=.8).add_legend()`
axes = fg.axes
axes[0,0].set_xlim(1,)
axes[0,0].set_ylim(1,)
fg.set(xscale="log", yscale="log")

[maahl]

Ah, sorry for the confusion! Thanks for the tips 😃