datetime to samplenumber slow

I have a very long vector with datetime and corresponding heart rate value. Now I want to calculate median value for every 5 minutes (my sample frequency is 4 seconds). However, sometimes some samples are missing. To be able to calculate the median value of HR every 5 minutes, I need to know which samples these are. I did it like this:

t1=time(1)

t2=time(end)

time5min=t1:minutes(5:t2)

for j=[time5min]

[~,ind1]=min(abs(datenum(time)-datenum(j-1)));

[~,ind2]=min(abs(datenum(time)-datenum(j)));

HR_med(ind2)=nanmedian(HR(ind1:ind2));

end

This is, as my vector is very large, very time consuming. Is there a faster way to do this?

Answer 1

[Steven Lord]

Do you want to compute the median for overlapping / sliding windows of 5 minutes, or do you want to compute it for each non-overlapping 5 minute window of your time period?

If you want overlapping bins, use movmedian specifying your datetime array as the SamplePoints and the K or NB/NF inputs as a duration of minutes(5).

If you want non-overlapping bins, take a look at the groupsummary function.

[Lieke Numan]

It is okay to have one median value for each 5 minutes, so no sliding window.

I have another function for which I need to use these timestamps. I have number of steps for each sample, and I want to calculate the number of steps per 5 minutes. So i take the difference in "steps(X)-steps(X-5minutes)".