Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: which variables are in the model?


From   Austin Nichols <austinnichols@gmail.com>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: which variables are in the model?
Date   Thu, 5 Feb 2009 13:46:12 -0500

Alan <Alan.H.Feiveson@nasa.gov>:
Why not refer to the column of e(b) by the name of the variable whose
coef you want?  Then you can trap the error that occurs if the column
does not exist, and decide what to do about it...

sysuse auto
qui reg pr mpg
mat b=e(b)
mat sub=b[1,"mpg"]
di sub[1,1]


On Thu, Feb 5, 2009 at 1:31 PM, Feiveson, Alan H. (JSC-SK311)
<Alan.H.Feiveson@nasa.gov> wrote:
> Hi - I have been running a self-programmed bootstrap resampling scheme using -bsample- with one of the xt-regression models. For each iteration of the bootstrap, I had been blissfully saving the coefficient estimates using e(b) which was supposed to have 11 components. Then I discovered that for some iterations, some model terms were dropped for collinearity, so for example, b[1,6] in one iteration may not refer to the same model term as b[1,6] in another iteration! Furthermore, if a current iteration had only 10 terms, then referring to b[1,11] does not produce an error, since b[1,11] contains the estimate from the last iteration in which all terms were used, etc. To do the correct bookkeeping, I need to know exactly which of the original set of 11 model terms were retained after each estimation. My question is how to know this? There doesn't appear to be anything in -ereturn- that gives me this information. It is true that e(rank) tells me how many terms were used, but i!
 t !
>  doesn't say which of the 11 terms were dropped (if any).

*
*   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/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index