It's a sliding window. From the text in the function:
% B = RMOUTLIERS(A,..., MOVMETHOD, WL) uses a moving window method to
% determine contextual outliers instead of global outliers. MOVMETHOD can
% be 'movmedian' or 'movmean'.
You'll probably have to write your own function to divide the vector into chunks like you want. It's easy enough with a for loop. Something like:
for idx = 1:100:numel(vec)/100
chunk = vec(idx:idx+100);
% chunk(isoutlier(chunk)) = nan; one option
% filtered(idx:idx+100) = chunk;
filtered(idx:idx+100) = filloutliers(chunk, 'linear');
end
