How to compare average value from recent 2 weeks to average of previous 6 weeks by group

I understood that your issue is that you don’t manage to get the rolling average based on the category “merchant”. If that is it, one way would be this one:

1. First groupby with resample Write the resample, the same way that you already had, I just added a ffill (fill down to avoid nans if you don’t have one of the datapoints):
   df.index = df.date
grouped = df.groupby(['merchant']).resample('W')[['amount']].mean()
grouped.ffill(axis = 0,inplace = True)

2. Apply groupby again, with rolling on the column “amount” for 2 and 6 weeks. Note that there will be nans in this example because for example to calculate the rolling of 2 weeks if will need at least the previous week:

grouped = grouped.reset_index()
grouped.index = grouped.date
mean_2 = grouped.groupby(["merchant"]).rolling(2).amount.mean()
mean_6 = grouped.groupby(["merchant"]).rolling(6).amount.mean()

3. Concatenate series
   result=pd.concat([mean_2,mean_6],axis=1)
result.columns = ["mean2","mean6"]

The result is like this: