Date:         Tue, 19 May 2009 06:29:41 -0500
Reply-To:     "Data _null_;" <iebupdte@GMAIL.COM>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         "Data _null_;" <iebupdte@GMAIL.COM>
Subject:      Re: variance covariance matrix
Comments: To: Umrao <umraoamit@gmail.com>
In-Reply-To:  <3bf42769-5971-4c06-919c-1786f3ee19b8@i28g2000prd.googlegroups.com>
Content-Type: text/plain; charset=ISO-8859-1

It's a bit fiddley but it works.  There are surely other methods that
I don't know.

data cov(type='COV');
   length _type_ _name_ $8;
   retain _type_ 'COV';
   input v1-v5;
   _name_ = cats('v',_n_);
   cards;
0.644607   0.1680246  1.483855    1.6281761 0.7427545
0.1680246  0.0626773  0.2708275   0.1894284 0.2474915
1.483855   0.2708275  4.8090391   5.3790562 1.8120638
1.6281761  0.1894284  5.3790562   8.7463271 0.6100965
0.7427545  0.2474915  1.8120638   0.6100965 2.2213662
;;;
   run;
proc factor noprint data=cov corr outstat=stats;
   run;
*proc print;
   run;
data corr;
   set stats;
   where _type_ eq 'CORR';
   array v[*] v:;
   do _n_ = 1 to dim(v);
      if v[_n_] = 1 then v[_n_] = .;
      end;
   run;
data covCorr;
   update cov corr;
   by _name_;
   run;
proc print;
   run;

On 5/19/09, Umrao <umraoamit@gmail.com> wrote:
> I have a variance-covariance matrix COV. The diagonal numbers are the
> variances and the off diagonal numbers are the covriances.
>
> COV={0.644607   0.1680246                      1.483855    1.6281761
> 0.7427545,
>         0.1680246      0.0626773       0.2708275          0.1894284    0.2474915,
>         1.483855       0.2708275                        4.8090391    5.3790562
> 1.8120638,
>           1.6281761    0.1894284                        5.3790562    8.7463271
> 0.6100965,
>            0.7427545   0.2474915                      1.8120638
> 0.6100965       2.2213662};
>
> Is there any procedure to convert the off diagonal numbers that are
> the covariances into correlations.
>
> i.e., for the second element int he first row i whant to calculate the
> value as  0.1680246/sqrt(0.644607*0.0626773)
> this i have to do for all the offdiagonal elements.
> The new matrix becomes a variance-correlation matrix as
>
> varcorr={0.644607       0.835930597     0.842781127     0.685710831     0.620708515,
> 0.835930597     0.0626773       0.493296992     0.255845094     0.663277199,
> 0.842781127     0.493296992     4.8090391                     0.829400075
> 0.554414079,
> 0.685710831     0.255845094     0.829400075     8.7463271
> 0.13841259,
> 0.620708515     0.663277199     0.554414079     0.13841259      2.2213662};
>
>
> I have to apply the formula for correlation= cov(x,y)/sqrt{var(x)*var
> (y)} to all the offdiagonal elements of the  variance-covariance
> matrix COV.
>