[FIXED] How to add calculations on a pandas groupby table?

Issue

I've created a table using pandas.groupby, but want to add a column that does a calculation on the columns in the groupby output. Here is my code (I'm using Python 2) and result:

DMM_pivot = df.query('rpc_flag != "IB RPC"').groupby(['rpc_flag', 'DMM_type'])
DMM_pivot = DMM_pivot['number_accounts'].sum().unstack('rpc_flag').fillna(0)
DMM_pivot = DMM_pivot.apply(lambda x: x / x.sum()).round(2)  # Show percentage of total
print DMM_pivot

rpc_flag  No RPC  OB RPC
DMM_type                
Bad         0.34    0.25
Good        0.66    0.75

I want to add a column that shows percentage difference accross the columns: (OB RPC value / No RPC value - 1) and expect these results:

rpc_flag  No RPC  OB RPC   Change
DMM_type                
Bad         0.34    0.25    -0.26
Good        0.66    0.75     0.14

Solution

Use apply.

DMM_pivot["Change"] = DMM_pivot.apply(lambda x: x["OB RPC"]/x["No RPC"] -1, axis=1)

Answered By - Igor Rivin