Description

List of model classes accepted by model.avg, model.sel, and dredge.

Details

Fitted model objects that can be used with model selection and model averaging functions include those produced by:

• lm, glm (package stats);

• rlm, glm.nb and polr (MASS);

• multinom (nnet);

• lme, gls (nlme);

• lmer, glmer (lme4);

• cpglm, cpglmm (cplm);

• gam, gamm* (mgcv);

• gamm4* (gamm4);

• gamlss (gamlss);

• glmmML (glmmML);

• glmmadmb (glmmADMB from R-Forge);

• glmmTMB (glmmTMB);

• MCMCglmm* (MCMCglmm);

• asreml (non-free commercial package asreml; allows only for REML comparisons);

• hurdle, zeroinfl (pscl);

• negbin, betabin (class "glimML"), package aod);

• aodml, aodql (aods3);

• betareg (betareg);

• brglm (brglm);

• *sarlm models, spautolm (spatialreg);

• spml* (if fitted by ML, splm);

• coxph, survreg (survival);

• coxme, lmekin (coxme);

• rq (quantreg);

• clm and clmm (ordinal);

• logistf (logistf);

• crunch*, pgls (caper);

• maxlike (maxlike);

• functions from package unmarked (within the class "unmarkedFit");

• mark and related functions (class mark from package RMark). Currently dredge can only manipulate formula element of the argument model.parameters, keeping its other elements intact.

Generalized Estimation Equation model implementations: geeglm from package geepack, gee from gee, geem from geeM, wgee from wgeesel, and yags from yags (on R-Forge) can be used with QIC as the selection criterion.

Other classes are also likely to be supported, in particular if they inherit from one of the above classes. In general, the models averaged with model.avg may belong to different types (e.g. glm and gam), provided they use the same data and response, and if it is valid to do so. This applies also to constructing model selection tables with model.sel.

Note

* In order to use gamm, gamm4, spml (> 1.0.0), crunch or MCMCglmm with dredge, an updateable wrapper for these functions should be created.

See Also

model.avg, model.sel and dredge.