How to group “remaining” results beyond Top N into “Others” with pandas

Question

When group a pandas dataframe by one column say \"version\" and which has 10 distinct versions. How can one plot the Top 3 (which cover over 90%) and put the small remainders in

Answer 1

I assume you also want the Other group to be summed, for your example to a total of 3?

If i was aiming to win the Pandas one-liner competition this would be my entry:

df.replace(df.groupby('Version').sum().sort('Value', ascending=False).index[2:], 'Other').groupby('Version').sum()

         Value
Version       
Other        3
Top1        19
Top2        13

But that's completely unreadable, so lets break it down:

You already showed how to sum each group, sorting this result and selecting anything outside of the top 2 can be done with:

not_top2 = df.groupby('Version').sum().sort('Value', ascending=False).index[2:]

In this example not_top2 contains Other1 and Other2.

We can replace those Versions to a common name with:

dfnew  = df.replace(not_top2, 'Other')
print dfnew

  Version  Value
0    Top1     14
1    Top1      3
2    Top1      2
3    Top2      6
4    Top2      7
5   Other      1
6   Other      2

The above replaces the contents of not_top2 in any column. A little substep is needed if you expect this value to occur in any other column than Version.

Whats left is to do your original grouping again:

dfnew.groupby('Version').sum()

Which gives:

         Value
Version       
Other        3
Top1        19
Top2        13