2
$\begingroup$

I'm not aware of any R package that lets me use specify the covariance pattern model such as in the package nlme and run the mixed effects poisson/logistic regression.

Is this because of mathematical limitations or just because I'm not aware of the package?

This isn't a package request because I'm requesting why it's not mathematically possible if it is not possible.

$\endgroup$
  • $\begingroup$ It is feasible. In SAS PROC GLIMMIX, there is option TOEP and AR(1). So search hard in R. I am not an export on R. $\endgroup$ – user158565 Aug 12 at 20:33
2
$\begingroup$

In GLMMs you do not have an analogue of multivariate error terms for which you can define such a correlation structure.

A potential way to achieve something like this in GLMMs would be to use observation-level random effects, and define such a correlation structure for their variance-covariance matrix. I think this should be provided by the glmmTMB package.

$\endgroup$
  • $\begingroup$ Does that meaning there's no protection against missing data as a result of specifying a correlation structure with mixed poisson/logit? $\endgroup$ – user252477 Aug 12 at 21:01
  • 1
    $\begingroup$ @Germania You do not necessarily need an AR1 or Toeplitz structure to be protected against missing data. You can also model the serial correlation by including random effects for time, i.e., random intercepts, linear random slopes of time, and nonlinear random slopes of time using polynomials or splines. $\endgroup$ – Dimitris Rizopoulos Aug 13 at 5:12

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy