ENH: Mosaic plot and DataArray

[scls19fr]

Hello,

I'd like to draw using Python a similar mosaic plot (also named Marimekko chart) of R dataset HairEyeColor.

https://github.com/wch/r-source/blob/trunk/src/library/datasets/data/HairEyeColor.R

Here is R code

> library(datasets)
> HairEyeColor
, , Sex = Male

       Eye
Hair    Brown Blue Hazel Green
  Black    32   11    10     3
  Brown    53   50    25    15
  Red      10   10     7     7
  Blond     3   30     5     8

, , Sex = Female

       Eye
Hair    Brown Blue Hazel Green
  Black    36    9     5     2
  Brown    66   34    29    14
  Red      16    7     7     7
  Blond     4   64     5     8

I can display a mosaic plot using R with the following example available thanks to ?HairEyeColor

require(graphics)
## Full mosaic
mosaicplot(HairEyeColor)
## Aggregate over sex (as in Snee's original data)
x <- apply(HairEyeColor, c(1, 2), sum)
x
mosaicplot(x, main = "Relation between hair and eye color")

I get:

I try to produce a similar plot using Python.

I created xarray.DataArray using:

import numpy as np
import xarray


data = np.array([32, 53, 10, 3, 11, 50, 10, 30, 10, 25, 7, 5, 3, 15, 7, 8,
                 36, 66, 16, 4,  9, 34,  7, 64,  5, 29, 7, 5, 2, 14, 7, 8])

_dim = (4, 4, 2)
data = data.reshape(_dim[::-1])

_dims = ['Hair', 'Eye', 'Sex']
_coords = [['Black', 'Brown', 'Red', 'Blond'], 
           ['Brown', 'Blue', 'Hazel', 'Green'],
           ['Male', 'Female']]

data = xarray.DataArray(
    data, dims=_dims[::-1],
    coords=_coords[::-1], name='Number'
)

assert int(data.loc['Female', 'Green', 'Black']) == 2

and try to plot mosaic using:

import matplotlib.pyplot as plt
from statsmodels.graphics.mosaicplot import mosaic
mosaic(data)
plt.show()

I get:

I think there is some room for improvements.

Providing a da.plot.mosaic method may be a first feature.
Using labels will be an other interesting feature.

Kind regards

PS: a similar mosaic plot can be plot using Titanic dataset

https://github.com/wch/r-source/blob/trunk/src/library/datasets/data/Titanic.R

with R

with Python

import numpy as np
import xarray

data = np.array([  0,   0,  35,   0,
                   0,   0,  17,   0,
                 118, 154, 387, 670,
                   4,  13,  89,   3,
                   5,  11,  13,   0,
                   1,  13,  14,   0,
                  57,  14,  75, 192,
                 140,  80,  76,  20])

_dim = (4, 2, 2, 2)
data = data.reshape(_dim[::-1])

_dims = ['Class', 'Sex', 'Age', 'Survived']
_coords = [['1st', '2nd', '3rd', 'Crew'],
           ['Male', 'Female'],
           ['Child', 'Adult'],
           ['No', 'Yes']]

data = xarray.DataArray(
    data, dims=_dims[::-1],
    coords=_coords[::-1], name='Number'
)

assert int(data.loc['Yes', 'Adult', 'Male', '3rd']) == 75

PS2: Python/xarray datasets are available at https://github.com/Rdatasets/python

PS3: mosaic code can be found here:
https://github.com/statsmodels/statsmodels/blob/master/statsmodels/graphics/mosaicplot.py
https://github.com/statsmodels/statsmodels/blob/master/statsmodels/graphics/tests/test_mosaicplot.py

[scls19fr]

Converting to Pandas Series with hierachical index seems to help:

mosaic(hair_eye_color.data.to_series())

mosaic(titanic.data.to_series())

It's not a clear as R mosaic plot... but that's not so bad.

What is your opinion about adding a mosaic plot method ?

[shoyer]

Interesting -- I haven't encountered mosaic plots before. If it's as simple as writing mosaic(data.to_series()) or data.to_series().pipe(mosaic), then I would lean against including this in xarray's API, if only because the wrapper wouldn't do very much, and mosaic plotting is (I suspect) a relatively rare operation for most analyses.

[scls19fr]

Mosaic plots are not so rare and widely used for categorical display.
I suggest reading
https://en.wikipedia.org/wiki/Mosaic_plot
https://www.perceptualedge.com/articles/visual_business_intelligence/are_mosaic_plots_worthwhile.pdf
http://www.theusrus.de/blog/understanding-mosaic-plots/
and http://www.datavis.ca/papers/moshist.pdf

[fmaussion]

I tend to agree with @shoyer that mosaic plots are probably a quite unusual way to represent DataArrays (but I can only speak for me).

Since the actual mosaic plot is done by statsmodels you should probably propose your suggestions for melioration there?