Search code examples
matlabmatrixmeanlarge-data

Matrix mean of values >x

I need to compute the mean of a matrix but only for those values that are greater than a specified number. Right now I come with this idea:

Media = mean(W_tot_migl,2);

H = W_tot_migl;
H(H<LimiteInferiore) = nan;
Media_b = nanmean(H,2);

Is there any way to avoid to create another matrix H = W_tot_migl?

Solution

  • For matrix H and threshold T you can use:

    M = nanmean(H+0./(H>T),2);

    in your case:

    Media_b = nanmean(W_tot_migl+0./(W_tot_migl>LimiteInferiore),2);

    OR calculate it yourself:

    M = sum(H.*(H>T),2)./sum(H>T,2);

    For your case:

    Media_b = sum(W_tot_migl.*(W_tot_migl>LimiteInferiore),2)./sum(W_tot_migl>LimiteInferiore,2);

    Note that both methods return NaN if there are no values greater than the threshold in a row.

    However, in general you can use accumarray to accomplish this:

    H2 = H>T;
[I,~] = find(H2);
M = accumarray(I, H(H2), [size(H,1) 1], @mean);

    This method returns zero if there are no elements larger than the threshold in a row.