Is there a pythonic way to do a contingency table in Pandas?

Question

Given a dataframe that looks like this:

            A   B      
2005-09-06  5  -2  
2005-09-07 -1   3  
2005-09-08  4   5 
2005-09-09 -8   2
2005-09-10 -2  -         


        

Answer 1

Probably easiest to just use the pandas function crosstab. Borrowing from Dyno Fu above:

import pandas as pd
from StringIO import StringIO
table = """dt          A   B
2005-09-06  5  -2
2005-09-07 -1   3
2005-09-08  4   5
2005-09-09 -8   2
2005-09-10 -2  -5
2005-09-11 -7   9
2005-09-12  2   8
2005-09-13  6  -5
2005-09-14  6  -5
"""
sio = StringIO(table)
df = pd.read_table(sio, sep=r"\s+", parse_dates=['dt'])
df.set_index("dt", inplace=True)

pd.crosstab(df.A > 0, df.B > 0)

Output:

B      False  True 
A                  
False      1      3
True       3      2

[2 rows x 2 columns]

Also the table is usable if you want to do a Fisher exact test with scipy.stats etc:

from scipy.stats import fisher_exact
tab = pd.crosstab(df.A > 0, df.B > 0)
fisher_exact(tab)