|
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]
Re: st: Problems with standard errors using xi3, svy and ologit
From |
[email protected] (Jeff Pitblado, StataCorp LP) |
To |
[email protected] |
Subject |
Re: st: Problems with standard errors using xi3, svy and ologit |
Date |
Thu, 14 Feb 2008 15:44:54 -0600 |
Iain Lang <[email protected]> four that -svy: ologit- is yielding different
standard error values when prefixed with -xi3- than with-xi-:
> I'm running models in Stata 9 using xi3, svy and ologit. This combination
> seems to produce very large standard errors, and these don't occur when I
> use xi instead of xi3. An example is pasted below (using a binary outcome as
> an example). Has anyone else encountered similar problems? Any help or
> comments appreciated!
Iain also reported Stata output showing the differences, which Richard
Williams confirmed. Michael Blasnik noticed that version control has
something to do with it.
Michael pointed out that
. xi3: version 9: svy: ologit ...
produces the same results as
. xi: svy: ologit ...
The problem is that -predict- is computing incorrect score values when
-ologit- was called under version control less than 9. The -svy- prefix fits
the model by calling -ologit- using the same version number it was called
under, then calls -predict, score- to get the equation-level scores in order
to compute the linearized variance estimates.
To clarify, -svy: ologit- (and -svy: oprobit-) only produced incorrect
standard errors when called under version control less than 9.
We just finished fixing this version control problem with -predict- after
-ologit- (and -oprobit-). It should be in the next ado-file update.
--Jeff
[email protected]
*
* 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/