[prev in list] [next in list] [prev in thread] [next in thread] 

List:       r-sig-mixed-models
Subject:    Re: [R-sig-ME] Please help on glmer in lme4
From:       Ben Bolker <bbolker () gmail ! com>
Date:       2018-06-15 19:05:10
Message-ID: b9c384fc-8d36-d09a-fec1-1ffaa06587cf () gmail ! com
[Download RAW message or body]


  Not quite, I think.  It's *much* more obscure than it should be, but I
believe that the initial GLM fixed-effects parameters in both stages are
set to a vector of zeros, not to the results of an initial GLM fit.  (In
the nAGQ=0 phase, the fixed effects parameters are profiled out, not
used as part of the high-level optimization.) The following two lines
are the proof (not at all obvious!)

https://github.com/lme4/lme4/blob/5071612b79bbc86ef7ae185249baf18c62afca56/R/modular.R#L430

https://github.com/lme4/lme4/blob/5071612b79bbc86ef7ae185249baf18c62afca56/R/AllClass.R#L80

On 2018-06-15 12:49 PM, lia moris wrote:
> Dear R,
> 
> 
> The stage 1 (nAGQ=0) use the initial values for theta by default and
> initial values for beta is obtained from a generalized linear model
> fit. The stage 2 initial values are  theta estimated from stage 1 and
> initial values for beta  the same ones from stage 1, right?
> 
> 
> 
> Thank's!
> 
> 
> Lia
> 
> 
> [[alternative HTML version deleted]]
> 
> _______________________________________________ 
> R-sig-mixed-models@r-project.org mailing list 
> https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
>

_______________________________________________
R-sig-mixed-models@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
[prev in list] [next in list] [prev in thread] [next in thread]