Question

I need your help in solving the following problem. Column1 shows a grouping by integer. Any non-nan value in Col2 should be inserted in the matrix for matching groupnumber(Col1).

mat = [ ...
        1  nan 
        1  0.1
        1  nan
        1  nan
        2  nan
        2  nan
        2  nan
        3  0.5
        4  nan
        4  nan
        4  nan
        5  0.2
        5  nan ] ;

ans = [ ...
        1  0.1 
        1  0.1
        1  0.1
        1  0.1
        2  nan
        2  nan
        2  nan
        3  0.5
        4  nan
        4  nan
        4  nan
        5  0.2
        5  0.2 ] ;

Please recommend a vectorized approach. Data is huge and is already being run in a for-loop. There will never be multiple non-nan values(col2) within a group (in mat). Thanks!

Answer 1

A solution using ACCUMARRAY will accomplish your goal:

values = accumarray(mat(:,1),mat(:,2),[],@min);
mat(:,2) = values(mat(:,1));

I use the function MIN here for convenience, since it will return the non-NaN value if there is one, or NaN if that’s all there is. It’s simpler than the logic involved in checking for any non-NaN values using, say, the function ISNAN. You could actually use the function MAX as well, since it behaves the same way in this case.