SVY - Mixed Effects Modeling in Complex Survey - Model Comparison Options

[Jenna Powell]

Hi all,

I am currently developing a mixed effects regression model with household as level 1 and cluster as level 2 for a two-phase, two-stage survey. I have STATA 17 running on Windows IOS. I am utilizing svy: meglm with a binomial family and logit link for my primary models. I have read through plenty of STATA17's svy estimation documentation and am finding little guidance for my question:

How can I compare nested models with the output provided by the svy: meglm command?

I am aware that Likelihood Ratio Testing and AIC/BIC are inappropriate for evaluating mixed effect models. Alternatively, Logit gof does not work in STATA 17. I aim to determine the most parsimonious model for a binary nutrition outcome.
A professor recommended I use the F-statistic and associated p-value to compare - if that is appropriate, how do I do that?

Thanks!

Model 1 output:
:

Model 2 output:

[Andrew Musau]

[SIZE=14px]
I am aware that Likelihood Ratio Testing and AIC/BIC are inappropriate

Wald tests are valid for complex survey data. So with nested models, you can still use the test command.

help test

[David Radwin]

I don't know if this will help you, but you might look at

. net describe st0099_1, from(http://www.stata-journal.com/software/sj10-2)

---------------------------------------------------------------------------------------------------------------
package st0099_1 from http://www.stata-journal.com/software/sj10-2
---------------------------------------------------------------------------------------------------------------

TITLE
      SJ10-2 st0099_1.  Update: Goodness-of-fit test for a...

DESCRIPTION/AUTHOR(S)
       Update: Goodness-of-fit test for a logistic regression
         model estimated using survey sample data
      by Kellie J. Archer, Ph.D., Department of Biostatistics,
           Virginia Commonwealth University
         Michael I. Lichter, Primary Care Research Institute,
           University of Buffalo
         Stanley Lemeshow, Ph.D., School of Public Health,
           The Ohio State University
      Support:  [email protected]
      After installation, type help svylogitgof

INSTALLATION FILES                             (type net install st0099_1)
      st0099_1/svylogitgof.ado
      st0099_1/svylogitgof.hlp

ANCILLARY FILES                                (type net get st0099_1)
      st0099_1/svylogitgof.do
      st0099_1/smallset.dta
      st0099_1/largeset.dta
      st0099_1/nhis2004.dta
---------------------------------------------------------------------------------------------------------------

which updates

Archer, K. J., & Lemeshow, S. (2006). Goodness-of-fit test for a logistic regression model fitted using survey sample data. The Stata Journal, 6(1), 97-105.