Pandas - Calculating daily differences relative to earliest value

问题

This is probably pretty easy, but for some reason I am finding it quite difficult to complete. Any tips would be greatly appreciated. I have some time series data consisting of 5-minute intervals each day, ala:

Date                   Values
2012-12-05 09:30:00    5
2012-12-05 09:35:00    7
2012-12-05 09:40:00    3
2012-12-05 09:45:00    2
2012-12-05 09:50:00    15
2012-12-06 09:30:00    4
2012-12-06 09:35:00    3
2012-12-06 09:40:00    8
2012-12-06 09:45:00    1

I would like to calculate the differences relative to the first value of the day (which in this case always will be the 9:30 value), ie. end up with this DataFrame:

Date                   Values
2012-12-05 09:30:00    0
2012-12-05 09:35:00    2
2012-12-05 09:40:00    -2
2012-12-05 09:45:00    -3
2012-12-05 09:50:00    10
2012-12-06 09:30:00    0
2012-12-06 09:35:00    -1
2012-12-06 09:40:00    4
2012-12-06 09:45:00    -3

回答1:

You need substract by Series created transform with groupby by Series.dt.date and first:

print (df.Values.groupby(df.Date.dt.day).transform('first'))
0    5
1    5
2    5
3    5
4    5
5    4
6    4
7    4
8    4
Name: Values, dtype: int64

df.Values = df.Values - df.Values.groupby(df.Date.dt.day).transform('first')

print (df)
                 Date  Values
0 2012-12-05 09:30:00       0
1 2012-12-05 09:35:00       2
2 2012-12-05 09:40:00      -2
3 2012-12-05 09:45:00      -3
4 2012-12-05 09:50:00      10
5 2012-12-06 09:30:00       0
6 2012-12-06 09:35:00      -1
7 2012-12-06 09:40:00       4
8 2012-12-06 09:45:00      -3

回答2:

You can use broadcasting:

df.Values - df.Values.iloc[0]

来源：https://stackoverflow.com/questions/40104449/pandas-calculating-daily-differences-relative-to-earliest-value