Could you send us the output of sessionInfo() please so we can see which version of the lme4 package you are using? In recent versions, especially the development version available as
install.packages("lme4", repos = "http://r-forge.r-project.org")
the PQL algorithm is no longer used. The Laplace approximation is now the default. The adaptive Gauss-Hermite quadrature (AGQ) approximation may be offered in the future.
If the documentation indicates that PQL is the default then that is a documentation error. With the currently available implementation of the direct optimization of the Laplace approximation to the log-likelihood for the model there is no purpose in offering PQL.
On Thu, Feb 14, 2008 at 6:50 PM, Daniel Malter <daniel_at_umd.edu> wrote:
> I run the following models:
> 1a. lmer(Y~X+(1|Subject),family=binomial(link="logit")) and
> 1b. lmer(Y~X+(1|Subject),family=binomial(link="logit"),method="PQL")
> Why does 1b produce results different from 1a? The reason why I am asking is
> that the help states that "PQL" is the default of GLMMs
> 2. gamm(Y~X,family=binomial(link="logit"),random=list(Subject=~1))
> The interesting thing about the example below is, that gamm is also supposed
> to fit by "PQL". Interestingly, however, the GAMM fit yields about the
> coefficient estimates of 1b. But the significance values of 1a. Any insight
> would be greatly appreciated.
> cuncta stricte discussurus
> R-help_at_r-project.org mailing list
> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
Archive maintained by Robert King, hosted by
the discipline of
statistics at the
University of Newcastle,
Archive generated by hypermail 2.2.0, at Fri 15 Feb 2008 - 16:30:14 GMT.
Mailing list information is available at https://stat.ethz.ch/mailman/listinfo/r-help. Please read the posting guide before posting to the list.
list of date sections of archive