Standard deviation ignoring 0s in matrix

Hi all, I have a 295x34 matrix called Absy. I want to take the standard deviation of all the columns, however some columns contain many zeros. I want to ignore the zeros and take the standard deviation (as zeros are only their due to some columns being different size).

Thank you.

Answer 1

[Matt Kindig]

One easy way would be to convert all zeros to NaN, and then use the nanstd function to ignore the NaNs in the std calculation.

Absy(Absy==0)=NaN;
s = nanstd(Absy,[],1);   %column by column std deviation

[Mate 2u]

Is their a way to do this with kurtosis also?

Answer 2

[Jan]

x = rand(295, 34);
x(rand(size(x)) < 0.1) = 0;

index = (x ~= 0);
sumnz = sum(index, 1);
meanx = sum(x, 1) ./ sumnz;
xm    = bsxfun(@minus, x, meanx);
xm(~index) = 0;
s     = sqrt(sum(xm .* xm, 1) ./ (sumnz - 1));

[NOT TESTED]