如何更改Seaborn对图中的回归模型

Question

I can create regression plot with seaborns regplot , where I can change the order of the linear regression model with the order option, or choose a lowess model with the lowess=True option as in:

sns.regplot(x='logAssets', y='logLTIFR', lowess=True, data=df, scatter_kws={'alpha':0.15}, line_kws={'color': 'red'})

and obtain this:

Is it possible to change the order of the linear regression in a pairplot ?

Or even better, to use a lowess model in seaborn pairplot ?

Answer 1

For more advanced usage, use PairGrid instead of pairplot . Basically, PairGrid allows you to control which function is used to plot the upper, lower and diagonal plots independently. Check out the documentation for PairGrid for more details.

To answer your specific question:

iris = sns.load_dataset("iris")
g = sns.PairGrid(iris)
g = g.map_upper(sns.regplot, lowess=True, scatter_kws={'alpha':0.15}, line_kws={'color': 'red'})