Average/mean of datetime hours and minutes ITSELF - not corresponding embedded data

How to I compute the mean/average time of a datetime array containing hours and minutes?

rowMin =

20×1 datetime array

09:00

09:10

09:00

09:14

09:59

09:05

09:00

09:35

13:15

09:00

09:00

15:00

09:35

10:15

14:10

16:35

09:00

11:15

09:00

10:20

How to I compute the mean/average time?

rowMin.Format='HH:mm'

mean(rowMin)

Does not work. Neither converting to table and subsequently timetable does:

rowMinTable=table(rowMin)

rowMinTimetable=table2timetable(rowMinTable)

dt = minutes(30);

output = retime(table2timetable(rowMinTable),'regular','linear','TimeStep',dt)

Also, converting to datenum to compute mean and then revert to datetime is not the solution, since it uses the dates as well (I am ONLY interested in the mean of the hours and minutes HH:MM):

rowMin.Format='HH:mm';

rowMinDatenum=datenum(rowMin)

rowMinDatenumMean=mean(rowMinDatenum)

rowMinMean=datetime(rowMinDatenumMean,'ConvertFrom','datenum')

This question is also asked on stack exchange:

https://stackoverflow.com/questions/65324040/average-mean-of-datetime-hours-and-minutes-itself-not-corresponding-embedded-d

Answer 1

[Cris LaPierre]

Changing the format does not change the underlying data. It just changes how it appears.

I think you would have to convert your datetimes to durations by subtracting the date, and compute the mean duration. Do the times all come from the same day?

[Stefan555]

The times comes from different days. But I have only shown the hours and minutes HH:MM because they are what I need as for the time being. If it can be computed from the original data, yyyy-mm-dd HH:MM then that would of course be a more general solution. But I don't need that.

As is apparent of the HH:MM shown in OP, the mean value is early in the morning, maybe around 10:30, so a computation of this value is what I am looking for.

[Cris LaPierre]

Ok, there are a few ways I can think of to do this. They share the same thing - you must extract the time component, then convert it to a duration. Otherwise, you'll have to convert you time to one unit (decimal days, hours or minutes).

This page might help give you some ideas.

Here's a dummy example:

% creating a data set. Ignore this part
d=(datetime(2020,1,1):days(1):datetime(2020,1,20))';
t = duration({'09:00';'09:10';'09:00';'09:14';'09:59';'09:05';'09:00';'09:35';'13:15';'09:00'; ...
   '09:00';'15:00';'09:35';'10:15';'14:10';'16:35';'09:00';'11:15';'09:00';'10:20'},'InputFormat','hh:mm');
% View as datetimes
D = d+t

D = 20×1 datetime array
   01-Jan-2020 09:00:00
   02-Jan-2020 09:10:00
   03-Jan-2020 09:00:00
   04-Jan-2020 09:14:00
   05-Jan-2020 09:59:00
   06-Jan-2020 09:05:00
   07-Jan-2020 09:00:00
   08-Jan-2020 09:35:00
   09-Jan-2020 13:15:00
   10-Jan-2020 09:00:00
   11-Jan-2020 09:00:00
   12-Jan-2020 15:00:00
   13-Jan-2020 09:35:00
   14-Jan-2020 10:15:00
   15-Jan-2020 14:10:00
   16-Jan-2020 16:35:00
   17-Jan-2020 09:00:00
   18-Jan-2020 11:15:00
   19-Jan-2020 09:00:00
   20-Jan-2020 10:20:00

% extract hour and minute using Hour and Minute properties, convert to duration
dur = duration(D.Hour, D.Minute,0,"Format",'hh:mm');
% Calculate the mean
mean(dur)

ans = duration
   10:31

[Eric Sofen]

A slight simplification. Instead of extracting the hours and minutes components, you can use the timeofday function to get durations that represent the time of day.