[R-sig-ME] nAGQ = 0

[Rolf Turner]

On 07/09/17 10:57, Poe, John wrote:

> This is where my being a political scientist on a listserv full of 
> definitely not political scientists is going to make me look dumb. I've 
> never actually seen a model specification like that before for a 
> multilevel model. Is the outcome supposed to be the proportion dead out 
> of the total population for each row? I'm missing something about this 
> that is probably very obvious.
> 
> After fiddling with it I was able to get it to converge for one chain 
> but I wouldn't trust it at all right now.

Well, I hadn't seen a model specification quite like that either.  It's 
brm() syntax, which is a bit different from glm() or glmer() syntax.

The model being fitted is a binomial model, with success probability p 
modelled by

     g(p) = beta_0 + beta_1 * x + Z_0 + Z_1 * x

where the Z_k are the random effects, and where g() is the link 
function, e.g. cloglog(). (I *think* I've got that right.)

Of course beta_0 and beta_1 depend on treatment group and Z_0 and Z_1
depend on "Rep", which is nested within treatment group.

Note that we are talking *generalized* linear mixed models here; the 
responses are binomial success counts.  "Success" = Dead, since one is 
trying to kill the bugs.

For what it's worth the glmer() syntax is

    glmer(cbind(Dead,Alive) ~ (0 + Trt)/x + (x | Rep),
          family=binomial(link="cloglog"),data=X)

I got the brm() syntax from the vignette; as I said, I'd never seen it 
before.

cheers,

Rolf

-- 
Technical Editor ANZJS
Department of Statistics
University of Auckland
Phone: +64-9-373-7599 ext. 88276