[R-sig-ME] "General" (non-Bernoulli) binomial models in GLMMadaptive.
d@r|zopou|o@ @end|ng |rom er@@mu@mc@n|
Sun Aug 4 14:28:11 CEST 2019
In general, the more quadrature points you use the better the approximation of the log-likelihood at the expense of computational time. The order of the approximation is improved every two quadrature points you add. Hence, you start at 1 (equivalent to Laplace approximation), and you go 3, 5, etc.
For more info check Section 5.3 of my course notes (http://www.drizopoulos.com/courses/EMC/CE08.pdf), and also this thesis: https://macsphere.mcmaster.ca/handle/11375/17272
From: Rolf Turner <r.turner using auckland.ac.nz<mailto:r.turner using auckland.ac.nz>>
Date: Sunday, 04 Aug 2019, 2:16 PM
To: D. Rizopoulos <d.rizopoulos using erasmusmc.nl<mailto:d.rizopoulos using erasmusmc.nl>>
Cc: R-mixed models mailing list <r-sig-mixed-models using r-project.org<mailto:r-sig-mixed-models using r-project.org>>
Subject: Re: "General" (non-Bernoulli) binomial models in GLMMadaptive.
On 4/08/19 10:10 PM, D. Rizopoulos wrote:
> The current CRAN version of GLMMadaptive should work for binomial data.
> For example, this run in my machine:
> system.time(fm1 <- glmer(cbind(incidence, size - incidence) ~ period + (1 | herd),
> data = cbpp, family = binomial, nAGQ = 21))
> system.time(gm1 <- mixed_model(cbind(incidence, size - incidence) ~ period, random = ~ 1 | herd,
> data = cbpp, family = binomial(), nAGQ = 21))
Thanks very much for this. And whew! That's a relief, since neither of
my proposed work-arounds seems to work worth a damn.
May I just ask a quick (said he, optimistically) follow-up question?
Can you provide a rationale for the choice of nAGQ = 21? (If this would
require a lengthy discourse, don't worry about it.)
P.S. I gather, from an off-list OOO response that I received, that
you are on a conference/vacation trip. My apologies for pestering you
under these circumstances. I hope that you are having an enjoyable time.
Honorary Research Fellow
Department of Statistics
University of Auckland
Phone: +64-9-373-7599 ext. 88276
[[alternative HTML version deleted]]
More information about the R-sig-mixed-models