for loop output into matrix for pairwise comparisons

I have 6 samples (mdio1-6), each with 100 observations that values range from 1 to 4 (integers). I would like to compare each of the samples for the number of observations that they are the same. So I have this code:

for n = 1:100
                A = hap.mdio1
                B = hap.mdio2
                dist(n) = sum(abs(A(n)-B(n)))
        end
        metric = (100-(sum(dist12(:)==0)))/100

I have a couple of questions.

1. I do not understand how to add a loop to go through each of the comparisons possible (i.e. mdio01-mdio3, mdio1-mdio4, mdio1-mdio5, mdio1-mdio6 mdio02-mdio03, mdio02-mdio4 etc.).

2. I would like to put the value calculated in 'metric' into a matrix so that it is a 6x6 with the metric values for each of the comparisons (would be zeros along the diagonal because they all match).

Thanks!

Answer 1

[James Tursa]

As you are starting to see, having variables with "dynamic" names ending in 1, 2, 3, etc. can quickly cause programming headaches. You are forced into duplicating and hand editing multiple loops to get the processing you need. Or you resort to using eval, making your code klunky, hard to read and harder to debug. And if you have more that a handful of such variables it becomes too much of a burden. There are much better ways to program than this. E.g., see these links:

http://www.mathworks.com/matlabcentral/answers/57445-faq-how-can-i-create-variables-a1-a2-a10-in-a-loop

http://www.mathworks.com/matlabcentral/answers/143-how-do-i-make-a-series-of-variables-a1-a2-a3-a10

[Pinkvirus]

Those two posts are actually where I started because I was trying to create the answers how it says not to in the post. Mdio1, 2 etc. are variables in a dataset, that were imported as vectors initially. It is now a dataset with 100 rows and 6 columns title "chr, pos, mdio1, mdio2 etc."

now my code reads:

 for n = 1:100
                A = hap.mdio1
                B = hap.mdio2
                dist(n) = sum(abs(A(n)-B(n)))
        end

        metric = (100-(sum(dist12(:)==0)))/100

Any advice on looping over all pairwise comparisons and depositing in a matrix?

Thanks!

Answer 2

[Stephen23]

https://www.mathworks.com/help/matlab/matlab_prog/string-evaluation.html

https://www.mathworks.com/matlabcentral/answers/304528-tutorial-why-variables-should-not-be-named-dynamically-eval

Answer 3

[Stephen23]

You can generate the fieldnames using sprintf:

>> str = sprintf('mdio%i',3)
str = 'mdio3'

And use this in a dynamic field name:

hap.(str)

will access the data corresponding to the field name given by str. You should preallocate the output array, and then simply allocate the values using indexing. But it might be simpler to use non-scalar structure instead, as you will see in my example solution below.

You might also be interested in using nchoosek to generate all combinations of the six samples:

>> X = nchoosek(1:6,2)
X =
   1     2
   1     3
   1     4
   1     5
   1     6
   2     3
   2     4
   2     5
   2     6
   3     4
   3     5
   3     6
   4     5
   4     6
   5     6

The difference can also be calculated on the whole vectors by using vectorized code, rather than calculating in a loop for each of the elements in the vectors.

All together you might get something like this (a simplified example):

A(4).data = [0,2,3,4,5];
A(3).data = [1,2,3,4,5];
A(2).data = [1,0,3,4,0];
A(1).data = [0,2,3,4,5];
B = nchoosek(1:numel(A),2);
D = zeros(numel(A));
for k = 1:size(B,1)
    x = B(k,1);
    y = B(k,2);
    D(x,y) = 1-mean(A(x).data==A(y).data);
end

where D is an upper-triangle matrix of the metric values:

>> D
D =
          0          0.6          0.2            0
          0            0          0.4          0.6
          0            0            0          0.2
          0            0            0            0