How do I CREATE a variance covariance matrix?

[Rosa Abraham]

Hi,

I'm estimating a model as given below

probit y z x1 x2 x3 x4
probit z y x1 x2 x5 x6

I've generated the residuals for each of these regressions
predict yhat
gen yresid = (y-yhat)

and similarly for z.

Now I want to get the variance-covariance matrix for yresid and zresid..

Is there anyway I can do this in Stata 10.

I hope this makes sense. Thank you in advance.

Regards,
Rosa Abraham

[Attaullah Shah]

Variance -covariance matrix can be estimated with user written command _^varrets

ssc install varrets

[Rosa Abraham]

Thank you Prof Shah. I had to install mvport
ssc install mvport
to get the varrets command..

Unfortunately, I'm using Stata 10, and varrets is a version 12.1 program. So I can't run this.. ! Any other options?

[Roman Mostazir]

Hi Rosa, not sure if I got you right. You have created two columns with residuals from two models. If you wish to see their correlation/covariance just type (you may have valid reasons for that),

corr yresid  zesid, cov

. But I intuitively think you are asking for the covariance matrix for the model you ran. This is how you can do it:

probit y z x1 x2 x3 x4

matlist e(V)   /*this will give you the variance/covariance matrix*/

/*Same can be obtained with the following*/

estat vce /*for var/cov matrix*/

estat vce, cor /*If you want them in correlation form rather in variance form*/

Note: The last option (cor) returns 1 for the diagonal elements.