Moving window of standard deviation

I am currently using movstd(A,k), and I want to detect values that is more than a certain threshold value at every array of local k-point standard deviation values. Is there any possible way that I can do that?

Answer 1

[Bjorn Gustavsson]

To do this it seems that it would be sensible to combine movstd and movmean (or movmedian). Then something like this ought to illustrate one way to do this:

% Mock up some trivial example-data
A = randn(31,1);
k = 7;
thresh = 0.1;
% Sliding statistics
mstdA = movstd(A,k);  % or movmad -    which measure of spread to use might depend on the
mavgA = movmean(A,k); % or movmedian,  distribution of the local "noise-variations"
% Detect points further from the moving average
iPos = find(A>(mavgA+mstdA)+thresh);
iNeg = find(A<(mavgA-mstdA)-thresh);
% Illustrating plots
plot(A,'.-')
hold on
plot(mavgA,'.-')  
plot(mavgA+mstdA,'.-')
plot(mavgA-mstdA,'.-')
plot(iPos,A(iPos),'ro')
plot(iNeg,A(iNeg),'bo')

HTH