maximum variable size allowed by the program is exceeded but it's not

There have been many many threads on this topic but none seem to apply to my case.

I'm running 64 bit matlab so that

[~,maxSize]=computer

returns

maxSize =   2.8147e+14

I'm working with a variable x which is a 10000 by 10000 array of scalars,

having approximate size 800000000, i.e., approx 1e-6 of maxSize.

I can compute mean(mean(x)), but when I try to compute

mean(mean(x(rows,cols))) where x(rows,cols)) is roughly a quarter the size of x,

I receive the dreaded error

Maximum variable size allowed by the program is exceeded.

even though, obviously, x(rows,cols) is minute, relative to maxSize.

Is there anyway to get around this problem?

thanks very much for any advice!

Leo

[Walter Roberson]

Are there possibly duplicates among rows or cols?

Does the same thing happen if you assign x(rows,cols) to a variable and then mean(mean()) ?

Have you considered using mean2() ?

[Leo Simon]

Thanks Walter. I discovered that when I reshaped my matrix into a vector, v, and computed mean(v(rows)) everything worked fine, and very quickly. Very odd, since the only change was a reshape, but since it worked, I'm happy.