How to create a matrix with the data of a comparison, upper triangular the percentage of similarity and lower triangular the times that were compared?

I have a matrix with N rows and 4 columns. Each row represent a comparison of a file between different sources, columns represent the sources, each of the elements of the matrix can take values form 1-4 or NaN.

Column 1 have values of 1 or NaN

Column 2 have values of 1,2 or NaN

Column 3 have values of 1,2,3 or NaN

Column 4 have values of 1,2,3,4 or NaN

An example is:

M  =  [1    NaN  3    3;
       1    2    NaN  2;
       NaN  2    2    4;
       1    2    1    1;
       1    1    3    1;
       1    1    3    4]

For example in row 1

[1    NaN  3    3]

sources 3 and 4 are similar, in source 2 there is not data, and source 1 is no similar to any other.

If I compare of M, source 1 with source 2 it was compared 4 times (because of NaN in M(1,2) and M(3,1)) and the similarity is 50% (2/4)

I would like to have a square matrix (4x4) with this data where the upper triangular matrix is the percentage of similarity and the lower triangular matrix the times that were compared, like this:

     0    50*   25    40
     4*    0    25    40
     4     4     0    40
     5     5     5     0

The stars are for the example when I compared source 1 with source 2 (indexes (1,2) and (2,1)).

I did that using a lot of IF and the script looks so ugly haha, "If" you can recomend me something I would really appreciate it

Answer 1

[Andrei Bobrov]

a = permute(M,[3,2,1]);
b = permute(M,[2,3,1]);
x = sum(bsxfun(@eq,a,b),3);
y = sum(~isnan(bsxfun(@plus,a,b)),3);
out = tril(y,-1) + triu(x./y'*100,1);

[German Preciat Gonzalez]

Dude, you are a genius!