how to ignore nan values in corr function?

sir, i have a three dimensional matrix say A(129*135*33). I have to find out the correlation coefficient along the third dimension with B=[1981:2013]. i am using the following code

        m1=squeeze(A(i,j,:));
        m2=[1981:2013]';
        p=[m1 m2];
        q=p(isfinite(p(:, 2)), :); %to remove the rows with nan values 
       CC(i,j)=corr(q(:,1), q(:,2);

due to some grid which have only nan values the program is not working and giving the following message

Error using corr (line 87) Requires a data matrix X.

please help how to solve it?

Answer 1

[Roger Stafford]

Clearly there is no sensible way to compute a correlation value for two empty vectors. I would suggest that for such cases you place a NaN in the CC matrix at the corresponding i,j pair:

   q = ...
   if numel(q) > 0
     CC(i,j) = corr(...)
   else
     CC(i,j) = NaN;
   end