如何计算 dataframe 每行的多个 Python Pandas datetime64[ns] 值的平均值？

Question

MC_schedule_df: 

    Act_Arr_Run-0        Act_Arr_Run-1          Act_Arr_Run-2        Act_Arr_Run-3
0   2005-08-05 05:15:08  2005-08-05 05:12:00    2005-08-05 05:16:50  2005-08-05 05:09:13
1   2005-08-05 06:18:30  2005-08-05 06:14:50    2005-08-05 06:14:29  2005-08-05 06:07:31
2   2005-08-05 06:22:17  2005-08-05 06:18:06    2005-08-05 06:26:25  2005-08-05 06:22:49
3   2005-08-05 08:52:56  2005-08-05 08:58:51    2005-08-05 09:05:27  2005-08-05 08:58:43
4   2005-08-05 13:04:24  2005-08-05 12:58:11    2005-08-05 13:05:41  2005-08-05 13:02:33
5   2005-08-05 13:22:08  2005-08-05 13:14:44    2005-08-05 13:09:08  2005-08-05 13:12:27
6   2005-08-05 14:26:38  2005-08-05 14:13:38    2005-08-05 14:17:31  2005-08-05 14:17:33
7   2005-08-05 18:08:41  2005-08-05 18:17:15    2005-08-05 18:14:21  2005-08-05 18:15:54
8   2005-08-05 19:46:15  2005-08-05 19:45:28    2005-08-05 19:46:20  2005-08-05 19:48:44
9   2005-08-05 23:13:53  2005-08-05 23:06:06    2005-08-05 23:06:25  2005-08-05 23:04:07

Hello,

I have the dataframe (MC_schedule_df) shown above, consisting of the following datatypes:

In[1]: MC_schedule_df.dtypes
Out[1]: 
Act_Arr_Run-0    datetime64[ns]
Act_Arr_Run-1    datetime64[ns]
Act_Arr_Run-2    datetime64[ns]
Act_Arr_Run-3    datetime64[ns]
dtype: object

The dataframe consists of rows of datetime values, of which i want to calculate the mean per row. I have tried the following code:

MC_schedule_df = MC_schedule_df.assign(Average=MC_schedule_df.mean(axis=1))

This results in a column filled with NaN values. I have tried to find out why this does not work and thus have read loads of documentation. My current guess is that Python is not able to 'destilate' the appropriate information from the datetime values to calculate the mean.

How to calculate the mean of these multiple Python Pandas datetime64[ns] values? Any help is appreciated.

Edit: i tried the methods of Datetime objects with pandas mean function . However, this method does not work, as i want to calculate the mean per row, and thus can not easily call the series.

Answer 1

You can use what shown in this answer . As pointed out in the link, you cannot calculate the mean of a bunch of dates, the operation is not supported. But you can calculate the average of a bunch of timedeltas.
Use the pandas apply function to generalize it and apply it to a DataFrame instead of a Series.

mean_values = MC_schedule_df.apply(lambda dt : (dt - dt.min()).mean() + dt.min(), axis=1)

Using your sample dataframe, mean_values is:

0   2005-08-05 05:13:17.750
1   2005-08-05 06:13:50.000
2   2005-08-05 06:22:24.250
3   2005-08-05 08:58:59.250
4   2005-08-05 13:02:42.250
5   2005-08-05 13:14:36.750
6   2005-08-05 14:18:50.000
7   2005-08-05 18:14:02.750
8   2005-08-05 19:46:41.750
9   2005-08-05 23:07:37.750
dtype: datetime64[ns]