Problems with gamma.mixed in Zelig
by Lester, Susan (Health)
Hi,
I'm using R 3.0.1, Zelig 4.1-3 and ZeligMultilevel 0.7-0
The model in the gamma.mixed demo gives false convergence:
library(ZeligMultilevel)
data(coalition2)
z.out <- zelig(
duration ~ invest + fract + polar + numst2 + crisis + tag(1 | country),
data = coalition2,
model = "gamma.mixed",
family = Gamma(link="log")
)
WARNING: Warning in mer_finalize(ans) : false convergence (8)
If I refit the model without any predictors, then it converges, but with pretty silly Std.Dev/Std.Error estimates - so this would seem to be where the problem lies:
Generalized linear mixed model fit by the Laplace approximation
Formula: duration ~ 1 + tag(1 | country)
Data: coalition2
AIC BIC logLik deviance
-1242560 -1242549 621283 -1242566
Random effects:
Groups Name Variance Std.Dev.
country (Intercept) 1.3621e+92 1.1671e+46
Residual 7.0281e+97 8.3833e+48
Number of obs: 304, groups: country, 14
Fixed effects:
Estimate Std. Error t value
(Intercept) 1.127e+02 4.808e+47 0