熊猫按两列分组，并从第三列输出值

Question

colour    num    accepted  returned
grey      1      yes       no
red       2      no        no
grey      4      yes       yes

I have the dataframe above and want to output unique combinations of colour and num columns and also the corresponding value in returned , as below

colour    num     returned
grey      1       no    
red       2       no      
grey      4       yes     

Using df.groupby(['colour', 'num']).size() gives me unique combinations but not the returned column.

Answer 1

If you're sure that the combination of colour and num is unique, you can just do:

df.groupby(['colour', 'num'])['returned'].max()

Of course, if it's not really unique and there is both a 'yes' and 'no' this will return 'yes' because 'yes' > 'no'...

But actually, this solution doesn't give anything else than df[['colour','num','returned']].drop_duplicates() , which is definitely leaner.

If somehow you know that rows can be repeated but the returned value is unique and you want at the same time the number of times it appears and the unique value in "returned", you can do it in one go with:

df.groupby(['colour','num'])['returned'].agg(['size','max'])

Which would return:

            size    max
colour  num     
grey    1   1       no
        4   1       yes
red     2   1       no

Answer 2

From your description, I think you should to group the returned column as well.

df.groupby(['colour','num','returned']).size() This will display the number of occurences of each returned status, grouped by num and colour :

colour  num  returned
grey    1    no          1
        4    yes         1
red     2    no          1