如何将value_counts应用于分组对象

Question

I have table that looks like this:

userid purchase_date
     1 2016-08-01
     1 2016-08-02
     2 2016-08-01
     2 2016-08-01
     3 2016-08-01
     3 2016-08-02
     3 2016-08-03

I am keeping track of each user's purchase history (a user can purchase multiple times a day). Now， I want to find the earliest date that the user made a purchase, so I did this:

df.groupby(userid).purchase_date.transform(min)

Now, I have the earliest purchase date for each. The next thing that I want to do is to apply value_count on it. so I expect to see this:

userid earliest_purchase_date
     1 2016-08-01
     2 2016-08-01
     3 2016-08-01

Apply value_counts on earliest_purcahse_date to get:

2016-08-01 3

How can I do that? I don't know what to do after the transformation.

PS I tried df.groupby(userid).purchase_date.transform(min).value_counts() this operation is performed on the entire df , not on each group.

Answer 1

I think you need groupby with idxmin for get indices of minimal values per group with selecting by loc :

print (df.groupby('userid')['purchase_date'].idxmin())
userid
1    0
2    2
3    4
Name: purchase_date, dtype: int64

df = df.loc[df.groupby('userid')['purchase_date'].idxmin()]
print (df)
   userid purchase_date
0       1    2016-08-01
2       2    2016-08-01
4       3    2016-08-01

And last value_counts :

print (df.purchase_date.value_counts())
2016-08-01    3
Name: purchase_date, dtype: int64

If need count how many minimal values is per group:

df = df.groupby('userid')['purchase_date']
       .apply(lambda x: pd.Series([len(x[x == x.min()]),x.min()],index=['count','min date']))
       .unstack()
print (df)

       count             min date
userid                           
1          1  2016-08-01 00:00:00
2          2  2016-08-01 00:00:00
3          1  2016-08-01 00:00:00