On Aug 20, 2004, at 2:33 AM, Oleksandr wrote:
I use matrix commands to estimate the coefficients of a regression. I
have got coefficient and variance/covarience matricies: beta and VC.
How
to get from VC to std. errors?
I suppose first line is
ER=vecdiag(VC)
The second step is to take a square root of each element of ER. How can
I do it in one line without going into loops and cyles?
More generally, how can I present the results in the form similar to
the
Stata: means, std. errors, t-statistics and conf. intervals?
If you write an e-class command, as described in the programming
manual, Stata will do all of this work for you. Here is a code fragment
that illustrates this. Prior to this point, I have used matrix commands
to tweak the b and V matrices to reflect a different specification:
* post the revised coefficient vector and v-c matrix, with
* the standard error of the transformed error term as an additional
entry
ereturn post `beta' `VV' , depname(`depn') obs(`enn') esample(`touse')
* display the results and return in ereturn
di _n "Scaled probit estimates (scale : {result:`scale'}) {col
51}Number of obs = {col 70}{result:{ralign 9:`enn'}}"
ereturn display
The 'ereturn display' will present the standard Stata estimation output
table, with standard errors, t or z values, p values, confidence
intervals, etc.
Just follow the rules in [p] on rolling your own estimation command.
Kit
*
* 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/
