# RE: st: How to estimate var-cov matrix from data

 From "Nick Cox" To Subject RE: st: How to estimate var-cov matrix from data Date Mon, 16 Feb 2004 11:34:18 -0000

```In addition,

. findit covariance matrix

points to an FAQ by Bill Gould,
a program -corrmat- by Shannon Driver,
a program -matcorr- by myself,
etc.

Nick
n.j.cox@durham.ac.uk

Richard Williams

wei liu

> >I have a very easy question to ask. I have two random
> >variables with 1000 observations each. How can I
> >generate the variance-covariance matrix for them? I
> >wanna save this matrix as a matrix variable for
> >further use. Is there a simple way to do so? Can I
> >further generalize this method for higher dimension?
>
> Try -help mataccum-
>
> The commands would be something like
>
> . mat accum cov = x y, noconstant deviations
> (obs=1000)
>
> . mat cov = cov/(_N-1)
>
> . mat list cov
>
> symmetric cov[2,2]
>             x          y
> x  9.0365155
> y  3.0502898   3.982194
>
>
> To confirm that you did it right,
>
> . corr x y, cov
> (obs=1000)
>
>               |        x        y
> -------------+------------------
>             x |  9.03652
>             y |  3.05029  3.98219
>
> You can do it with more than 2 variables; and you can make
> things more
> complicated by using if, in and weights parameters.

*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/
```