How to do a nanmean not including zero values.

Hello,

I have some awkward sea ice data, where I need to do some averaging of just the none zero values.

Land values are NAN and I would like to keep them as such, and ocean values are zeros. I would like to do some monthly averaging between years, which I would normally do as-

siz = size(A) ;
 B = zeros(siz(1), siz(2), siz(3), 12) ;
 for mId = 1 : 12
    B(:,:,:,mId) = nanmean(A(:,:,:,mId:12:end), 4) ;
 end 

Normally I would just set all zero values to NAN for this, however, I need to keep them as zero for my file to work further down the line.

Thank you,

Holly

[Guillaume]

" I would like to do some monthly averaging between years"

For this, I would strongly recommend storing your data into a table or timetable. Assuming one of the table or timetable variable is Date, then getting the monthly average is as simple as

monthlyaverage = groupsummary(yourtable, 'Date', 'monthly', 'mean')

I'm not entirely clear on what you're asking but if you want to ignore 0s in the monthly mean it's a simple modification of the above.

If you need help converting your data into a table then explain how it is currently stored.

Note that there's been very little reason to use the nanmean function from the stats toolbox for a long time now. The mean function from base matlab has been able to ignore NaNs since 2014, IIRC.

[J. Alex Lee]

It does sound like just ignoring zeros...so if the tabular storage works for you, instead of

'mean'

I guess you'd want to use something like

@(x)mean(x(x~=0),2,'omitnan')

where dimension 2 assumes you're operating a table column

I agree it would be nice if your data can be converted to the more meaningful timetable and operations look more human readable...

...But to the original question, I might be interpreting way too simplistically, but just in case, is it as simple as making a copy of A to modify for purposes of averaging, and using the original A downstream?

[Guillaume]

While we're at it, note that the original code can be simplified to

B = mean(reshape(A, size(A, 1), size(A, 2), size(A, 3), 12, []), 5, 'omitnan');

Answer 1

[Matt J]

Normally I would just set all zero values to NAN for this, however, I need to keep them as zero for my file to work further down the line.

I don't see what prevents you from simply modifying a copy of A instead of A itself,

siz = size(A) ;

Atmp=reshape(A,siz(1), siz(2), siz(3), 12,[]);
Atmp(A==0)=nan;

B=mean(Atmp,5,'omitnan');