MAD outlier analysis in for loop

Question

I'm running median absolute deviation outlier analysis on a matrix B (126 x 7). I thought my code below was working but I get an error about dimension mismatch. Can anyone help me? I'm still a beginner with MATLAB.

b = B; %identifying raw data for outlier analysis
k=size(B,2); %preallocating size of loop to run to be equivalent to number columns in B


for j= 1:k
    a=b(:,j); %identify data used
    fa = abs(a-median(a))./(mad(a,1)./.6745); %if x>3.5 then outlier
    dec_mad=fa>2.24; %logical operator identifying outlier
    nout(j)=sum(dec_mad); %sum of outliers
    x = dec_mad ==0; % logical of data with outliers removed
    b(:,j) = a(x); %data without outliers
end

Answer 1

This line is the issue:

b(:,j) = a(x);

If x contains some zeros (i.e. outliers were found), the output of a(x) is smaller than your original a. Therefore you are trying to replace b(:,j) - 126 values - with a(x) - less than 126 values, hence dimension mismatch.

You could just replace any outliers with NaN:

b(dec_mad,j)=NaN;

Or use a cell array which would allow your individual output columns to be different sizes.

out{j} = a(x);

Incidentally, it's best not to use j for loops, it's a MATLAB built-in for complex numbers.