Deleting row/column iteratively of a correlation matrix

Hi guys,

So I have a correlation matrix E which is quite large. Some of the individual (i,j) entries have E(i,j)>0.95 (or similar high values). I want to remove the j row/column in the correlation matrix since the correlations of j can be explained by i anyway (i and j are highly correlated). I want to do this for all such (i,j) pairs (this will have to be done iteratively for each entry where E(i,j)>0.95). However, I would also like a record of the j removed from the original matrix. For example:

E= 1     0.97     0.4
   0.97  1        0.38
   0.4   0.38     1   

Hence I want the algorithm to select E(1,2) entry, and delete the second row and column. It will then move to the second such E(i,j)>0.95 and record the j removed ( but I would like to know what column/row j that has been removed from the original matrix). The output of the code will be the reduced matrix and the column/rows j deleted from the original matrix E.

Thank you so much for all your help!

Answer 1

[Matt J]

I see no need to do it iteratively. It can be done as a one-step process.

 remove=any(E>0.95);

 E=E(~remove,~remove);

[Ansh]

Hi, yes I think I was severely over complicating a very simple problem. Sorry for the late acceptance and thank you for your help. I also forgot to add that I have taken out the trivial diagonal 1 cases (correlation matrix)