Logical Indexing issue

Hi,

I use Matlab mostly to handle and analyze large environmental datasets. Recently I've been trying to reduce matrices with NaNs in them to a subset. This would be easy if a given row, matrix(i,:), was all NaNs, but only one of the elements in the third column of that row is a NaN. Given that, I'd like to reject that row, and establish a second matrix without that/those row(s). I'm trying the following at the moment:

    reject=zeros(length(matrix1.(:,1)),1);    
    len=length(matrix1(:,1));
    for j=1:len
       reject(j,1)=isfinite(matrix1(j,3));
    end
    reject3=logical(horzcat(reject, reject, reject));
    matrix2=matrix1(reject3);

In theory, this should scan the third column of matrix1 (where NaNs are possible), and establishe a logical of that column, then make a [lenx3] logical of that which matches matrix1 in dimensions. Matrix2, then, should be the reduced set. But when I do this, matrix2 is returned to me as a vector of the three columns combined.

Can you suggest a way to go about this, or a quick fix to my code that might make this possible? Feel free to point out any other problems in the code. I'm a musician-turned-inefficient Matlab user... I have a feeling this could be achieved in four simple lines if I knew what I was doing.

TIA, Jonathan

Answer 1

[Jonathan]

Ah, perfect! I knew someone out there would know how to do that simply (or could think it up on the fly). Thanks very much, Jan. That worked brilliantly.

Jonathan

[Fangjun Jiang]

You can add your acknowledgement to the comment of the answer. Please delete this entry and accept Jan's answer.

Answer 2

[Jan]

I'm not sure, but perhaps you want this:

index   = isfinite(matrix1(:, 3));
matrix2 = reshape(matrix1(cat(2, index, index, index)), [], 3);

Easier:

matrix2 = matrix1(isfinite(matrix1(:, 3)), :);