[R-sig-ME] models with overdispersion and autocorr.

[Emmanuel Charpentier]

Le vendredi 22 mai 2009 à 11:45 +0100, Alison Johnston a écrit :
> Hi there
> 
> I'm trying to fit repeated count models at several locations.  The locations are a random effect as there are >50, and we're not interested in the actual location values.
> But the data needs to be fitted with a quasi or zero-inflated model, and there is autocorrelation through time.  
> 
> I can't find a function which allows quasi/zi AND autocorrelation to be fitted.  Is there one?  Or is there another way the model could be constructed to avoid the problem?

Douglas Bates has written the necessary correlation and covariance
functions ... for nlme (function lme and such...). These functions do
not (yet) exist in lme4 (lmer and consorts), alas ... You might try to
bug him, but what he said about his current schedule makes me strongly
doubt this attemp would be met with any success (or popularity with him,
BTW :)...

I don't know what GAMs can do in this case, and I don't know what has
been implemented in availagle R packages.

May I drag from my (cluttered) cave the suggestion to find a relevant
change of variable which, temporarily disccarding the random effects,
might allow an approximate analysis with ... lme ? In other words, can
you think of a variable transformation that would approximately
normalize your residuals ? In this case, correlation and variance
modeling functions available in nlme would take care of autocorrelation
and possible remaining heteroscedasticity. Then lme and/or its nonlinear
cousins would be able to account for your random effects.

HTH,

					Emmanuel Charpentier