Fwd: Question on Multi-collinearity

[boisvert@nist.gov]

Sender: Sione <sionep@xtra.co.nz>
Subject: Question on Multi-collinearity




Dear JAMA users,

I just want to ask this quick question of testing if multi-collinearity 
in a matrix dataset exist or not and also how to eliminate it?  Here is 
of what I thought the test would be:

--------------------------------------
double[][] a = new double[][]{{...}};
Matrix A = new Matrix(a);
int cols    = A.getColumnDimension();

int rank   =  A.rank();
if(rank<cols){
  throw new Exception("Dataset contains columns which exhibits 
multi-collinearity")
 }
--------------------------------------

I am not sure whether the code snippets shown above is the right 
approach of how to test multi-collinearity? 

My other question is, if the test on the data is shown that 
multi-collinearity exist (such as above) , then is there any way of   
identifying   columns which if  they are removed, this eliminates 
multi-collinearity. Is this possible to do it in JAMA?

Thanks,
Sione.


----- End forwarded message -----