st: Problem recreating SE and CI using e(V) from logistic, stcox etc

["Lee, William" <william.lee@kcl.ac.uk>]

Dear Statalisters,
 
I am wanting to extract figures from a regression model, but find large differences between the SE as calculated from the variance covariance matrix and that which appears in the output. First using stcox, and since replicated in logistic. What is odd is that the CIs (calculated from my SE) agree to many decimal places with each other. Simple example using auto, below:
 
sysuse auto, clear
logistic foreign mpg
qui {
matrix B=e(b)
matrix V=e(V)
matlist B
matlist V
local or=exp(B[1,1])
local se=sqrt(V[1,1])
local cihi=exp(B[1,1]+1.96*`se')
local cilo=exp(B[1,1]-1.96*`se')
noi dis "OR=`or' SE=`se' 95%CI=`cilo' `cihi'"
}

I have found Maarten Buis' "Where did my p-values go?", and a short statalist correspondence here:

http://www.stata.com/statalist/archive/2009-01/msg01090.html
 
The correspondence seems to be about the same issue but it does not appear to be resolved.

Can anyone know of an explanation for this?

Thanks for your consideration,

Will

Dr William Lee BSc MBChB MRCPsych MSc,
MRC Training Fellow,
General Hospital Psychiatry,
Department of Psychological Medicine,
Institute of Psychiatry,
Weston Education Centre,
10 Cutcombe Rd,
London,
SE5 9RJ 


*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/