standard deviation of non-zero elements of columns of a matrix

Hello,

How can I find the standard deviation of each column of a matrix ignoring the zero values? I must add that I do not have the stats package so I cannot use nanstd! Thank you!

Answer 1

[the cyclist]

Here is a straightforward way.

% Some input data
x = rand(7,6);
x(1,1) = 0; % Put in a couple zeros by hand
x(6,5) = 0;

% Preallocate the array that will hold the standard deviations
sd = zeros(1,6);

% Calculate and store the standard deviations of the non-zero elements.
for nc = 1:6
  indexToNonZero = x(:,nc)~=0;
  sd(nc) = std(x(indexToNonZero,nc));
end

[osminbas]

Thank you very much!

Answer 2

[Walter Roberson]

There is a MATLAB File Exchange contribution to provide these functions.

Answer 3

[David Freese]

A way to go about it using just matrix operations:

% Some input data
x = rand(7,6);
x(1,1) = 0; % Put in a couple zeros by hand
x(6,5) = 0;

% Calculate and store the standard deviations of the non-zero elements.
% using the identity std^2 = Var(x) = mean(x.^2,2) - mean(x,2).^2
avg = sum(x,2)./sum(x~=0,2);
sd = sqrt(sum(x.^2,2)./sum(x~=0,2) - avg.^2);

% Zero out any of the columns that were all zeros
sd(isnan(sd)) = 0;