Rearrange a correlation matrix into a vector and keep track of the corr names.

Hi All,

I have a large correlation matrix and I need to reshape the upper triangular (or lower triangular) information into a vector, which I can do as

ind = tril(true(size(Corr)),-1)
CorrVec  = Corr(ind)

What I also need is to easily know the nature of the correlation when I will iterate over the vector: suppose

var1 var2 var3

Corr = var1 [ a b c

vare2 b e f

vare3 c f i ]

(a=e=i=1)

I want to get what is below (I m not interested on how the rearrangement is done, I just need the 3 info to have a one-to-one correspondence )

CorrVec = [ b c f ];

CorrVarFirst = ['var1' , 'var1','var2']

CorrVarSecond = [ 'var2', 'var3', 'var3']

In this way I can easily deduce that CorrVec(2) [c] is the correlation between CorrVarFirst(2) [var1] and CorrVarSecond(2) [var3]

Thanks in advance!

Answer 1

[Turlough Hughes]

You can get the corresponding row and column indices as follows:

[iRow, iCol] = find(ind);

Then you can get the corresponding variable names as follows:

params = ["var1", "var2", "var3"];
disp(params(iRow))
disp(params(iCol))