groupby + apply + lambda fail to call my function

Question

I wrote a function that should normalise values belonging to same week (dividing by the value of the first day):

def normalize(week):
        norm_week = (week / week[0]) -1
        return norm_week

I get the week data from a groupby call and I pass it to the normalise function through the apply method (with lambda):

dataset['col_1_norm'] = 
dataset.groupby('week_number')['col_1'].apply(lambda x: normalize(x)) 

This is input dataset:

week_number   col_1 
week_1.       300 
week_1        500
.....         ... 
week_2        350
.....         ...

I expect the normalised values in the column "col_1_norm", but python returns multiple errors. (example -> 3361 return self._engine.get_loc(casted_key))

Where am I wrong ?? Could you pls help ? Thanks Charlie

Answer 1

I inserted a print statement in your normalize routine to debug:

def normalize(week):                                                            
    print(week)                                                                 
    norm_week = (week/week[0]) -1                                          
    return norm_week 

Here's what printed out:

0      300                                                                                               
1      500                                                                                               
2      900                                                                                               
3     1300                                                                                               
4      200                                                                                               
5      400                                                                                               
6      100                                                                                               
7      800                                                                                               
8      500                                                                                               
9      600                                                                                               
10     500                                                                                               
Name: week_1, dtype: int64                                                                               
11     800                                                                                               
12    1700                
13    1800                                                                                               
14    2000                                                                                               
15    1500                                          
Name: week_2, dtype: int64                                                                               
Traceback (most recent call last):
...

There's no index 0 in the week_2 group. Perhaps you meant to use iloc?

def normalize(week):                                                            
    norm_week = (week/week.iloc[0]) -1                                          
    return norm_week