how can propagate error for division of two matrices?

I have two arrays E(i,j,k) and E_2(i,j,k) in which some elements of these matrices are nan values. while, I can calculate standard deviations S(i,j) = nanstd(E(i,j,:)); and S_2(i,j) = nanstd(E_2(i,j,:)); for E and E_3, i don't know howe to calculate standard deviation for E_2(i,j,k)./E(i,j,k). i was wondering if it is possible using propegation of error (function sigma = PropError(f,varlist,vals,errs)) or any other way.

Thank you in advance

Answer 1

[Arvind Narayanan]

Hi Joseph,

If the second matrix(E(i,j,k) in this case) has all non zero elements(NaN allowed), then only we can calculate standard deviation of this p./q type matrix. Otherwise, the resulting matrix will have Inf in it which would result in standard deviation in NaN .

Since Inf makes the mean and other central tendencies of the matrix non-quantifiable, the result is NaN.

For matrix with non-zero elements, we can use p./q and then use nanstd to find standard deviation.

Hope this helps,

Arvind

[Joseph]

These two matrices are measurements with the known error for each element. lets say for matrix E(i,j,k) there is an error(i,j,k) matrix. and same for E_2 matrix. I didn't mention this in my question.