Question on covariance matrix loop

Really basic question from a beginner- I have a 60,000 by 4 matrix (AB) and want to calculate 2000 covariance matrices ( every 30 rows down, and all four columns).

Do you guys know whats wrong with my logic? The below sample loops the 2nd and third covariance matrices.

for i = 1:2 j=i*31; k= j+30; VarCov(i) = cov(AB( j:k ,:)); end

I get an error message of:

 In an assignment  A(I) = B, the number of elements in B and
 I must be the same.

But doesnt varcov consists of 2 matrices??

Answer 1

[Shashank Prasanna]

If you want to stack 2000 cov matrices then you may want to use 3D matrices,

 VarCov(:,:,i) = cov(AB(j:k,:));

Here for each i you will have a 4x4 cov matrix

[Shashank Prasanna]

mean(s,3) will work if you store it this way

Answer 2

[Azzi Abdelmalek]

a=rand(60000,4);
b=permute(reshape(a',4,30,[]),[2 1 3]);
for k=1:size(b,3)
  s{k,1}=cov(b(:,:,k));
end
out=cell2mat(s);

[Mike]

Thanks!- This works great. If I want to create an average of the 2000 matrices (stored as "s"), does mean(s,3) work?

[Azzi Abdelmalek]

What is the size of the average result?