Difference to group mean in a pandas data frame?

Question

Lets's assume I count how many oranges (Orange) and apples (Apple) people (id) eat in a certain time period. I also know if they are young or old (group). The pandas dataframe would maybe look like this:

df = pd.DataFrame({'id' : ['1','2','3','7'],
                   'group' : ['Young', 'Young', 'Old', 'Old'],
                       'Apple' : [7,2,5,4],
                       'Orange' : [3,6,4,4],
                       })

We can easily compute means using groupby(). E. g:

df.Apple.groupby(df.group).mean()

outputs

Old      4.5
Young    4.5

But let's say, I want to find how much the amount of apples and oranges consumed differs to the group mean per individual?

That is, the output should be

df = pd.DataFrame({'id' : ['1','2','3','7'],
                   'group' : ['Young', 'Young', 'Old', 'Old'],
                       'Apple' : [7,2,5,4],
                       'Orange' : [3,6,4,4],
                       'Apple Difference' : [2.5, -2.5, 0.5, -0.5],
                       })

Is there a way to do this with pandas/numpy? Sorry for the rockie-question Best /R

Answer 1

You need transform for mean with same length as df and substract by sub:

print (df.groupby('group')['Apple'].transform('mean'))
0    4.5
1    4.5
2    4.5
3    4.5
Name: Apple, dtype: float64

df = pd.DataFrame({'id' : ['1','2','3','7'],
                   'group' : ['Young', 'Young', 'Old', 'Old'],
                       'Apple' : [7,2,5,4],
                       'Orange' : [3,6,4,4],
                       })
df['Apple Difference'] = df['Apple'].sub(df.groupby('group')['Apple'].transform('mean'))
print (df)
   Apple  Orange  group id  Apple Difference
0      7       3  Young  1               2.5
1      2       6  Young  2              -2.5
2      5       4    Old  3               0.5
3      4       4    Old  7              -0.5