[R-sig-ME] Justification to exclude random effect

[Reuben Smit]

I'm modeling presence/absence using binary logistic regression. The
complexity begins with using a model-averaged logistic mixed model to
predict onto a new and very large dataset. I'm having trouble with
non-conformable argument error messages, only with the logistic regression.
I can model counts with the random effect and those average and predict
fine with the same dataset. Standard glm predicts very well. I suppose I
should provide a reproducible example, since the error is really why I'm
trying to avoid the random effect, and that maybe content for a separate
post.

-Reuben

On Mon, May 9, 2016 at 5:36 PM, Baldwin, Jim -FS <jbaldwin at fs.fed.us> wrote:

> I'm not seeing enough information to make such a decision - although in
> any event, I don't see that the slight additional complexity of a single
> random effect would warrant the need to re-run the analysis without it.
> Plus, if the random effect was part of the experimental layout, then (other
> than maybe having too few levels to be able to estimate adequately it as a
> random effect) it ought to be in the model.
>
> Does this deal with count data or something more continuous?  Are the
> residuals well behaved?
>
> Jim
>
>
> -----Original Message-----
> From: R-sig-mixed-models [mailto:r-sig-mixed-models-bounces at r-project.org]
> On Behalf Of Reuben Smit
> Sent: Monday, May 09, 2016 4:16 PM
> To: r-sig-mixed-models at r-project.org
> Subject: [R-sig-ME] Justification to exclude random effect
>
> Hello,
>
> The results of my mixed model show the single random effect variance and
> standard deviation equal to zero. A statistician reviewed my results and
> suggested the random effect is not explaining any additional variance apart
> from the fixed effects. I'm asking the list to support or refute this claim
> based on the aforementioned variance and standard deviation estimate. If
> indeed the random effect is not contributing to the model, is it justified
> to exclude the random effect and may I use a more simple generalized linear
> model with fixed effects only?
>
> Also, the random effect is the effect of Site from a random nested study
> design. My conjecture is that a longitudinal/spatial fixed effect is
> accounting for relatedness between sample points located in the same Site
> (block).
>
> Thanks in advance,
> Reuben
>
> [[alternative HTML version deleted]]
>
> _______________________________________________
> R-sig-mixed-models at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
>
>
>
>
> This electronic message contains information generated by the USDA solely
> for the intended recipients. Any unauthorized interception of this message
> or the use or disclosure of the information it contains may violate the law
> and subject the violator to civil or criminal penalties. If you believe you
> have received this message in error, please notify the sender and delete
> the email immediately.
>

	[[alternative HTML version deleted]]