Stacked bar plot by grouped data with pandas

Question

Let's assume I have pandas dataframe which has many features and I am interested in two. I'll call them feature1 and feature2.

feature1 can have three possible values. feature2 can have two possible values.

I need bar plot grouped by feature1 and stacked by count of rows with each value of feature2. (So that there will be three stacks each with two bars).

How to achieve this?

At the moment I have

import pandas as pd
df = pd.read_csv('data.csv')
df['feature1'][df['feature2'] == 0].value_counts().plot(kind='bar',label='0')
df['feature1'][df['feature2'] == 1].value_counts().plot(kind='bar',label='1')

but that is not what I actually want because it doesn't stack them.

What about `stacked` argument in [plot](http://pandas.pydata.org/pandas-docs/stable/generated/pandas.DataFrame.plot.html)? — Anton Protopopov, Jan 21 '16 at 07:40
isn't it what you want ? http://stackoverflow.com/questions/22787209/how-to-have-clusters-of-stacked-bars-with-python-pandas — jrjc, Jan 21 '16 at 08:41
@jeanrjc Sort of, but not exactly. I needed plot such as in given answer. — justanothercoder, Jan 21 '16 at 08:45

score 29 · Answer 1 · edited May 23 '17 at 12:15

29

Also, I have found another way to do this (with pandas):

df.groupby(['feature1', 'feature2']).size().unstack().plot(kind='bar', stacked=True)

Source: making a stacked barchart in pandas

edited May 23 '17 at 12:15

Community

1
1

answered Jan 21 '16 at 08:49

justanothercoder

1,830
1
16
27

1

What does the `size()` and `unstack()` do? Thanks. – tommy.carstensen Nov 23 '17 at 14:38
simple and clean! – jxn Feb 13 '18 at 01:56

score 5 · Accepted Answer · answered Jan 21 '16 at 08:07

Im not sure how to do this in matplotlib (pandas default plotting library), but if you are willing to try a different plotting library, it is quite easy to do it with Bokeh.

Here is an example

import pandas as pd
from bokeh.charts import Bar, output_file, show
x = pd.DataFrame({"gender": ["m","f","m","f","m","f"],
                  "enrolments": [500,20,100,342,54,47],
                  "class": ["comp-sci", "comp-sci",
                            "psych", "psych",
                            "history", "history"]})

bar = Bar(x, values='enrolments', label='class', stack='gender',
         title="Number of students enrolled per class",
         legend='top_right',bar_width=1.0)
output_file("myPlot.html")
show(bar)

`bokeh.charts` is deprecated and removed now. Do you have any alternative for this? — Ibrahim Gharyali, Apr 28 '21 at 15:20

score 2 · Answer 3 · answered Jan 18 '19 at 17:22

size produces a column with a simple row count for that grouping, its what produces the values for the y axis. unstack produces the row and column information necessary for matplotlib to create the stacked bar graph.

Essentially it takes

>>> s
one  a   1.0
     b   2.0
two  a   3.0
     b   4.0

and produces:

>>> s.unstack(level=-1)
     a   b
one  1.0  2.0
two  3.0  4.0

Stacked bar plot by grouped data with pandas

3 Answers3