Issue with PP-plot and different distributions

[erykml]

• Python version: Python 3.6.8
• numpy version: 1.14.3
• matplotlib version: 2.0.2
• mpl-probscale version: 0.2.3
• Operating System: MacOS Mojave 10.14.3

Description

I tried modifying the examples from the documentation and created two PP-plots: one using Standard Normal Distribution as the theoretical distribution, another one using N(100, 5). And both plots look exactly the same (this is not true for QQ-plots). Am I missing something?

What I Did

import warnings
warnings.simplefilter('ignore')

import numpy
from matplotlib import pyplot
import seaborn
from scipy import stats
import probscale
clear_bkgd = {'axes.facecolor':'none', 'figure.facecolor':'none'}
seaborn.set(style='ticks', context='talk', color_codes=True, rc=clear_bkgd)

# load up some example data from the seaborn package
tips = seaborn.load_dataset("tips")

%matplotlib inline
%config InlineBackend.figure_format ='retina'

common_opts = dict(
    plottype='pp',
    probax='x',
    datascale='log',
    datalabel='Total Bill (USD)',
    scatter_kws=dict(marker='+', linestyle='none', mew=1)
)

norm = stats.norm(100, 5)

fig, (ax1, ax2) = pyplot.subplots(figsize=(10, 6), ncols=2, sharex=True)
fig = probscale.probplot(tips['total_bill'], ax=ax1, dist=norm,
                         problabel='N(100, 5) Probabilities', **common_opts)

fig = probscale.probplot(tips['total_bill'], ax=ax2, dist=None,
                         problabel='Standard Normal Probabilities', **common_opts)

seaborn.despine()

[phobson]

Thanks for the precisely described and demonstrated issue.

I think this stems from a gap in the documentation. A PP plot doesn't actually use the distribution of the data. It's displaying the percentiles of the data, which are unaffected by any inferred or provided distribution.