RE: st: comparing nested models after multiple imputation

[Kieran McCaul <kieran.mccaul@uwa.edu.au>]

...

For multilevel models, have a look at the REALCOM Impute package at http://missingdata.lshtm.ac.uk/



-----Original Message-----
From: owner-statalist@hsphsun2.harvard.edu [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of Chelsea Garneau
Sent: Monday, 21 May 2012 9:09 AM
To: statalist@hsphsun2.harvard.edu
Subject: Re: st: comparing nested models after multiple imputation

The problem is that there is no likelihood ratio test after multiple
imputation.  All I have to go on is an F statistic for each
model.....could I just compare with a chi-square test, maybe?  Not
sure what the df would be......but I've read that the F statistic is
supposed to be comparable to the Wald chi-square test.....so maybe it
would just be 1 df?

New Question:  I'm also trying to figure out how to calculate my own
beta's for multilevel linear regressions....this option is also not
available in Stata from what I've read so far.  I know that betas are
supposed to be the coefficient X (standard error of the independent
variable/ standard error of the dependent variable).  But where I do I
find the S.E. of the dependent (this might be a really silly
question).  I can obviously get the standard deviation of the variable
overall, but is that the same as the S.E. in the regression?

On Sun, May 20, 2012 at 7:14 PM, Brendan Halpin <brendan.halpin@ul.ie> wrote:
> On Sun, May 20 2012, Chelsea Garneau wrote:
>
>> I'm trying to find a way to compare nested models using xtreg, xtlogit,
>> xtmelogit, xtmixed after doing multiple imputation.  By "nested models" I
>> mean subsequent models with additional predictors added, not multilevel -
>> though my data are multilevel data as well. Because the LL's are not
>> pooled, there is no e(ll) for the lrtest.  Rubin's rules combine parameter
>> estimates and standard errors, not LL's, but what is the best way to test
>> nested models using xt commands after mi?
>
> That's a very interesting question[1]. Would it be sufficient to run
> parallel LR-tests for each imputation, and report the mean or median
> p-value, or the count of times H0 is rejected?
>
> Brendan
>
> [1] By which I mean, I hope someone else has a more useful answer to it.
> --
> Brendan Halpin,   Department of Sociology,   University of Limerick,   Ireland
> Tel: w +353-61-213147  f +353-61-202569  h +353-61-338562;  Room F1-009 x 3147
> mailto:brendan.halpin@ul.ie    ULSociology on Facebook: http://on.fb.me/fjIK9t
> http://teaching.sociology.ul.ie/bhalpin/wordpress         twitter:@ULSociology
>
> *
> *   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/

*
*   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/

*
*   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/