pandas_value_counts_with_percent.md

Problem

• I want to use pandas.Series.value_counts() or pandas.DataFrame.groupby([a_col]).size() to get frequency counts; but I also want the percentages in one go.
• pandas.Series.value_counts() has a normalize=True option for percentages; I want this to be neatly merged with the counts with clear column labels.

Solution

• method chain with concatenation of 2 series to combine the 2 value counts (counts and percents)
• rename axis and set_axis to adjust new column names inside the chain
• pipe with lambdas to re-format counts (comma separator) and percents (from a decimal to percent)
• TODO: make a groupby version of this to support multiple groupings with a similar count and percent column output

import pandas as pd
import seaborn as sns

def value_counts_with_percent(a_series):
    counts_df = (    
        pd.concat([a_series.value_counts(dropna=False),
                   a_series.value_counts(dropna=False, normalize=True)
                  ]
                  , axis='columns'
                 ) # end concat
        .rename_axis(a_series.name)
        .set_axis(['count', 'pct'], axis='columns')    
        .pipe(lambda _df: _df.assign(count=_df['count'].apply(lambda s: '{:,}'.format(s)),
                                     pct=_df['pct'].apply(lambda s: '{:.2f}%'.format(s*100)),
                                    ) # end assign
             ) # end pipe
        .reset_index()
    )

    return counts_df

df = sns.load_dataset('mpg')

print(df['name'].pipe(value_counts_with_percent)[:10].to_markdown())
print()
print(df['origin'].pipe(value_counts_with_percent)[:10].to_markdown())

Output

	name	count	pct
0	ford pinto	6	1.51%
1	toyota corolla	5	1.26%
2	amc matador	5	1.26%
3	ford maverick	5	1.26%
4	chevrolet chevette	4	1.01%
5	amc gremlin	4	1.01%
6	chevrolet impala	4	1.01%
7	peugeot 504	4	1.01%
8	amc hornet	4	1.01%
9	toyota corona	4	1.01%

	origin	count	pct
0	usa	249	62.56%
1	japan	79	19.85%
2	europe	70	17.59%