Diagnostic tests for correlated random effects model

[Riaz Chaus]

Hi,

What diagnostic tests can be done on a correlated random effects model? And if so how do I code this in Stata? I am using panel data and there are time-varying and time-invariant variables in my model.

Many thanks,
Riaz

[Joseph Coveney]

Maybe something like the following for one example?

version 15.1

clear *

set seed `=strreverse("1495719")'

quietly drawnorm intercept slope, double corr(1 -0.5 \ -0.5 1) n(200)
generate int pid = _n

quietly expand 5
bysort pid: generate byte tim = _n
generate byte trt = mod(_n, 2)

generate double rsp = intercept + ((tim - 3) / 2 / 10) * (1 + slope) + rnormal(0, sqrt(0.5))

mixed rsp i.trt c.tim || pid: tim, covariance(unstructured) nolrtest nolog

*
* Begin here
*
predict double fitted, fitted
predict double residuals, residuals
graph twoway scatter residuals fitted, mcolor(black) msize(vsmall) ///
    xlabel(-3(1)3) ///
    ylabel(-3(1)3, angle(horizontal) nogrid) yline(0, lcolor(black) lpattern(dash))

exit

Maybe out-of-sample predictions for the fixed effects for another?

What is it that you hope to diagnose?

[Riaz Chaus]

Thank you for that Joseph, though I am not sure it is exactly what I need. I am trying to diagnose if the correlated random effects model that I have used is a good fit for my data. For example with a fixed effects model you can run an F-test to see if variables are jointly significant, is there such a thing for a CRM model. Sorry I am not so knowledgeable with econometric theory so maybe I haven't explained this well!

Many thanks,
Riaz

[Phil Bromiley]

See also
Schunck R. 2013. Within and between estimates in random-effects models: Advantages and drawbacks of correlated random effects and hybrid models. Stata Journal 13(1): 65-76.
Schunck R, Perales F. 2017. Within- and between-cluster effects in generalized linear mixed models: A discussion of approaches and the xthybrid command. Stata Journal 17(1): 89115.