Software notes

R

R packages:

• lme4 (glmer)
  • very full-featured in terms of random effects structure (nested and crossed, applied to any parameter), offsets, etc., but still in rapid development and not thoroughly documented. Laplace and AGQ
• MASS (glmmPQL)
  • PQL only, but very flexible (front end for nlme random effects, so allows spatial, temporal correlation structures etc.)
• glmmML
  • AGQ of random-intercept (single random effect) models, some bootstrap capability
• glmmADMB
  • Laplace of various models; includes negative binomial. Rough edges.
• glmmAK
• MCMCglmm
• glmmBUGS
  • generates BUGS model files from model specification. Nested random effects only?
• RLRsim
  • restricted likelihood tests of variance components by randomization
• hglm
  • Generalized linear models with random effects (based on the h-likelihood). Estimation with hierarchical generalized linear models

SAS

• PROC GLIMMIX
  • previously only PQL, but now (version 9.2) does Laplace and AGQ as well
• PROC NLMIXED
  • very flexible in terms of models etc., but multiple/crossed random effects tricky
• PROC GENMOD (repeated statement)

Genstat

• asREML
• HGLM methods?

others

• MLWin, ?
• WinBUGS etc.