How can I skip zero values in an array when evaluating the correlation coefficient ?

Hello, I am kind of new to Matlab and I am running into some issues evaluating the correlation coefficient between the natural logarithm of two arrays(chan1 and chan2) each of size 577*1 that includes zero values that I would like to skip when calculating the correlation coefficient. I thought I could do a nested for loop to skip the zero values and this is what I have:

for chan1 ~= 0
    for chan2 ~=0
c = corrcoef(log(chan1),log(chan2))
    end
end

But it seems like that Matlab does not like it when one uses does not equal to in a for loop

Please help me solving this issue.

Thanks!

[Faisal Alghamdi]

Another thing I thought of doing is:

if chan1 == 0
    return
elseif chan2 == 0
        return
else c = corrcoef(log(chan1),log(chan2))
end

but this is what I get:

c =

     NaN   NaN
     NaN   NaN

Answer 1

[the cyclist]

% Some pretend data
chan1 = round(randn(1000,1));
chan2 = round(randn(1000,1));

% Identify pairs where both are nonzero
idxBothNonzero = chan1 > 0 & chan2 > 0; 

% Evaluate correlation, using only pairs where both are nonzero
c = corrcoef(log(chan1(idxBothNonzero)),log(chan2(idxBothNonzero)))

[Faisal Alghamdi]

It works, Thank you very much!!

[Jonas Bauer]

Hi, can you suggest a solution for a matrix with several "0" cells, which should be ignored for corr?