Comparing multidimensional array with 2-D array

I have an array with indices A(1388, 56, 720). I want to do a nanmean along the third dimension. If that was all, B = nanmean(A,3) would be fine. However, I actually only want to get a nanmean when the number of non-nan values in the third dimension are greater than 360 (or other number). I would like to avoid doing some kind of series of if statements going element by element through A.

I can find the number of non-nan elements easily at each point in the first two dimensions by:

C = sum(~isnan(A),3);

However, I can't figure out how to make that into something easily useful. For instance,

D = nanmean(A(C > 360),3);

produces a vector in D of less than 1388*56 points with no indication of where these should be in the A(1388,56) array.

Any help to make something out of this would be very much appreciated.

Answer 1

[Ameer Hamza]

Something like this

A = rand(1388, 56, 720);
A(A<0.5) = nan; % making 50% elements nan
Aa = A; % make a copy of A

mask = repmat(sum(isnan(A), 3) < 360, 1, 1, size(A,3));
A(mask) = nan; % setting all the 3rd column values to nan so that nanmean ignore it.
B = nanmean(A,3); % nan indicate that this row column were ignored in 3rd dimension
B(isnan(B)) = 0;

[Carol Clayson]

Thank you very much! The key line here was the mask = repmat . . .

My only question was why you made Aa

[Ameer Hamza]

In case, you need to use the value of A again later in code. If it is not required, then you can overwrite it.