LanguageR/ lme4: any help on p-values for models with random correlation parameters?

Question

I use the package languageR for mixed effect models with the syntax at the end of this posting. I can use pvals.fnc to get p -values for models 1 and 3 ( hd_lmer1 and hd_lmer2 ). Using this with model two gives the following error message:

p2 = pvals.fnc(hd_lmer2) Error in pvals.fnc(hd_lmer2) : MCMC sampling is not yet implemented in lme4_0.999375 for models with random correlation parameters

I would be grateful if any one could help me out on how to get p-values for such models.

Models:

hd_lmer1 <- lmer(
  rot ~ time + group + sex + gen + (1 | subject) + (1 | rot.pre),
  data = data_long,
  REML = TRUE
)

hd_lmer2 <- lmer(
  rot ~ time + group + sex + gen + (time | subject) + (1 | rot.pre),
  data = data_long,
  REML = TRUE
)

hd_lmer3 <- lmer(
  rot ~ time * group + sex + gen + (1 | subject) + (1 | rot.pre),
  data = data_long,
  REML = TRUE
)

Answer 1

It is an old post but here is one possible solution that can be helpful, using a model comparison method to test if hd_lmer2 produces a better fit than hdlmer1 (ie, if the addition of the random effect is significative or not).

hdlmer1ml<-update(hdlmer1,REML=FALSE)
hdlmer2ml<-update(hdlmer2,REML=FALSE)

anova(hdlmer2ml,hdlmer1ml)