Suppose we have something like this:
df = pd.DataFrame([[0,1,0,13], [1,0,1,14], [1,1,0,12], [1,0,0,15]], columns = ["A", "B" , "C", "p"])
A, B, C have binary values and I want to compute mean of p for each column, but for each group (1 and 0) separately.
For one column I use
df.groupby('A')['p'].mean()
But how to compute mean for columns ABC at once?
Use DataFrame.melt for unpivot first, aggregate mean and last add Series.unstack:
df1 = df.melt('p').groupby(['variable', 'value'])['p'].mean().unstack(fill_value=0)
Or use DataFrame.pivot_table:
df1 = df.melt('p').pivot_table(index='variable', columns='value', values='p', fill_value=0)
print (df1)
value 0 1
variable
A 13.000000 13.666667
B 14.500000 12.500000
C 13.333333 14.000000