[R-sig-ME] Question on HGLM package HGLM2 function

[Ben Bolker]

Do you have only a single observation per group (union)? This looks like
the kind of error message you would get if the random effect were
confounded with the residual variance (which wouldn't be a problem with a
Poisson response)

On Sat, Jan 28, 2023, 8:08 PM J.D. Haltigan <jhaltiga using gmail.com> wrote:

> Hi:
>
> Perhaps someone on the list may be able to assist me in troubleshooting why
> when I run the following model using HGLM2:
>
> >GLM2_1 = hglm2(posXsymp~ treatment+proper_mask_base+prop_resp_ill_base_2 +
> (1 | union), family=gaussian(link=identity), rand.family=Beta(link=logit),
> maxit = 100, data = my_data)
>
> I get the following error:
>
> # Error in glm.fit(x = c(1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,  :
> #                          NA/NaN/Inf in 'x'
> #                        In addition: Warning message:
> #                          In hglm.default(X = X, y = Y, Z = Z, family =
> family, rand.family = rand.family,  :
> #                                            Residuals numerically 0 are
> replaced by 1e-8
>
> Yet, when I run a similar model:
>
> HGLM2_5 = hglm2(posXsymp~ treatment+proper_mask_base+prop_resp_ill_base_2 +
> (1 | union), family =poisson (link = log), rand.family =Gamma(link=log),
> data = my_data)
>
> I have no problems.
>
> The only difference, of course, between the two models is one is specified
> as a Gaussian run and one is specified as a poisson run. The data are the
> same, variables are the same etc.
>
> The Y response distribution in both cases is binomial (symptoms or not),
> but the error in the first run seems to suggest my issue is with the X
> variable response distributions.
>
> Any insights would be most appreciated.
>
> Best,
> J.D.
>
>         [[alternative HTML version deleted]]
>
> _______________________________________________
> R-sig-mixed-models using r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
>

	[[alternative HTML version deleted]]