Plotting a Categorical Variable in matplotlib with pandas

plotting_categorical_variables.py

"""

Plotting a categorical variable
----------------------------------
`df` is a pandas dataframe with a timeseries index. 
`df` has a column `categorical` of dtype object, strings and nans, which is a categorical variable representing events

----------------------------------
>>> print df[:5]
                        categorical
date                           
2014-03-15 14:56:50       users
2014-03-15 14:56:50       users
2014-03-15 14:57:15       users
2014-03-15 14:56:56      photos
2014-03-15 14:57:10      photos

>>> print "Type:", df.categorical.dtype
Type: object

>>> print "Unique vals: ", df.categorical.unique()
Unique vals:  [users photos streams nan]

"""

# 1. We want to plot `df.categorical` to see the distribution of events over time
figure() 

# create a mapping from categorical values to unique integers:
integer_map = dict([(val, i) for i, val in enumerate(set(df.categorical))]) 

# plot using this integer map, applied over `categorical`
ax = df.categorical.apply(lambda x: integer_map[x]).plot(
    marker='.',linestyle='', alpha=0.2,
    )

# Set y tick locations
from matplotlib.ticker import FixedLocator
fl = FixedLocator(range(len(integer_map)))
ax.yaxis.set_major_locator(fl)

# Set y tick labels
ax.yaxis.set_ticklabels([str(k) for k in integer_map.viewkeys()])

# 2. We want to see a histogram of value counts
figure()

df.categorical.value_counts().plot(kind='bar')
title('Histogram of events by event type')