Find correlation between two matrices with varying NaN coordinates

Hello. I have global two matrices, A and B

They are both the same dimensions, but A has NaN's in some places that B does not. When I use corrcoef, it returns NaN's, even though the values within A and B are not identical. Is it returning NaN because the varying coordinates of NaN values within each matrix?

Does anyone know what I'm doing wrong in this case?

Thanks,

Melissa

Answer 1

[Chad Greene]

Indeed,

A = [1 2 3 4 5 6 7]; 
B = [1 2 2 4 5 6 NaN]; 
corrcoef(A,B)
ans =
     1   NaN
   NaN   NaN

But if you consider only non-NaN data:

% indices of non-nan data:
realz = isfinite(A) & isfinite(B); 

% correlation coefficient of non-nan data:
corrcoef(A(realz),B(realz))
ans =
    1.0000    0.9786
    0.9786    1.0000