Remove rows or cols whose elements are all NaN

Question

0 个投票

I have a matrix A:

A = [1   2   3   4 NaN;
     NaN NaN NaN NaN NaN;
     1   2   NaN 4 5;
     1   NaN 3   4 5];

and what I just want to is:

A = [1   2   3   4 NaN;
     1   2   NaN 4 5;
     1   NaN 3   4 5];

Currently I'm using the follwoing:

method = 1;
A = A'; % transposing is just for convenience
if method == 1
  A = A(:, ~isnan(nanmean(A))); % nanmean throws NaN if all elements are NaN
elseif method == 2
  A = A(:, sum(~isnan(A)) ~= 0);
end
A = A'; % re-transpose

However, I think nanmean is for statistical purpose and not for matrix manipulation. Moreover, actually I have much much much larger matrix. Thus, such operation might require longer computational time.

And I don't think all(isnan(A)) or all(~isnan(A)) is appropriate in some special cases.