[R-sig-eco] CLMM and model selection conflict

[Law, Jason]

Chris,

I'm no expert here, but you can obtain LRT's using anova() with the models you want to test. I don't see a reference to using drop1 with clmm in any of the package docs. But there is for clm which makes me think this may be intentional on the part of the package authors.

I would think that all of the dangers of testing fixed effects in the presence of random effects is present here just as they are with lmer. In that case, a parametric bootstrap or mcmc seems like the preferred way to go (http://glmm.wikidot.com/faq).

Jason

-----Original Message-----
From: R-sig-ecology [mailto:r-sig-ecology-bounces at r-project.org] On Behalf Of Chris.Wilcox at csiro.au
Sent: Friday, June 05, 2015 3:11 PM
To: r-sig-ecology at r-project.org
Subject: [R-sig-eco] CLMM and model selection conflict

Hi all,

I am using CLMM in package ordinal to model the condition of animal carcasses (on a 1 � 5 scale of decay) in an experiment evaluating the relationship between time since death and condition.  I am fitting models to a small number of covariates (5) in addition to the time.  We followed the condition of individual animal carcasses over time, in a variety of experimental conditions (on land, floating, submerged; high/low current, high/low temperature; fresh/frozen).  I want to do some model selection on the models.  I have tried add1/drop1 from the ordinal package, along with dredge from MuMIn.  In both cases the functions are not able to find a formula from the fitted model object.  After much searching, I have not found any others in the various R help mailing lists that have run into the same problem.  Any suggestions would be appreciated.

Thanks,

Chris

Chris Wilcox
Principal Research Scientist
CSIRO Marine and Atmospheric Research
Tel +61 3 6232 5306
Mob +61 4 3939 7317
www.csiro.au/Marine-debris
http://people.csiro.au/W/C/Chris-Wilcox/
http://www.imas.utas.edu.au/qms/staff-profiles/csiro-staff-profiles/dr-chris-wilcox2

	[[alternative HTML version deleted]]