[R] Mixed-effects model for pre-post randomization design

[Ben Bolker]

Marco Barbàra <jabbba <at> gmail.com> writes:

> 
> DeaR userRs,
> 
> I recently read this Liang-Zeger article:
> 
> http://sankhya.isical.ac.in/search/62b1/fpaper7.html
> 
> in which (among other things) they adopt a random intercept model for
> pre-post designed trials, using a conditional likelihood approach
> (I didn't think it possible with only two measurements per subject)
> 
> I'm trying to figure out (if and) how it is possible to reproduce
> straightforwardly their model using R standard mixed model tools, but
> I cannot even try to reproduce their work, since they used a
> non-available dataset (I found an extract on prof. Diggle's web site
> where it is explicitly reported to be "confidential"), so I have to
> review a bit of likelihood theory along with some implementation
> details.
> 
> In the meantime, I wonder if anyone here could point out any related
> documentation to me.
> 


  This might get more attention on r-sig-mixed-models at r-project.org.
I took a quick look at the paper, but it's not a case where the
answer is immediately obvious.  The paper of reference for lme4
(see http://cran.r-project.org/web/packages/lme4/citation.html )
gives technical details of lme4's implementation, in case that's
useful.