glmm: Fit Generalized Linear Mixed-Effects Models Using State Space...

In helske/StateSpaceGLMM: Inference of Generalized Linear Mixed Models Using State Space Representation.

Description

Function glmm estimates GLMM using methods basec on state space modelling.

Usage

glmm(response.var, grouping.var, fixed.formula, random.formula, data,
  distribution, init.random, init.dispersion, init.fixed,
  correlating.effects = TRUE, REML = TRUE, estimate.dispersion,
  return.model = TRUE, nsim = 0, maxiter = 50, maxeval = 1000,
  xtol_rel = 1e-06, ...)

Arguments

response.var	Name of the response variable in data.
grouping.var	Name of the grouping variable in data. Only one grouping variable is allowed.
data	Optional data frame environment containing the variables in the model.
distribution	Character vector defining the distributions used for each group. Either length of one (same distribution for all groups) or length of p, number of groups. Default is "gaussian".
init.random	Initial values for random effect covariances.
init.dispersion	Initial values for dispersion paremeters.
init.fixed	Initial values for fixed effects.
correlating.effects	Logical. Default is TRUE.
estimate.dispersion	Logical. Is the dispersion parameter estimated or fixed.
return.model	Logical. Is the estimated model returned in output.
nsim	Integer. Number of independent samples used in importance sampling. Default is 0, which corresponds to Laplace approximation.
maxiter	Integer. Number of iterations for in iterative weighted least squares.
method	"REML", "REML-ML" or "ML". "REML-ML" uses estimates obtained from REML for ML estimation. Note that ML is much slower.

Examples

data(butterfly)

require(lme4)
system.time(fit.SSglmm.reml<-glmm(response.var="Colias", grouping.var="site", fixed=~habitat + building + urbanveg,random=~1,
                                 REML=TRUE,data=butterfly,distribution="poisson",return.model=TRUE))

system.time(fit.SSglmm.ml<-glmm(response.var="Colias", grouping.var="site", fixed=~habitat + building + urbanveg,random=~1,
                               REML=FALSE,data=butterfly,distribution="poisson",return.model=TRUE))

fit.glmer<-glmer(Colias ~ habitat + building + urbanveg + (1|site),
                family=poisson, data=butterfly,control=glmerControl(optimizer = "bobyqa"))

sqrt(fit.SSglmm.reml$random$P)
sqrt(fit.SSglmm.ml$random$P)
VarCorr(fit.glmer)

fixef(fit.glmer)
fit.SSglmm.ml$fixed$coef
fit.SSglmm.reml$fixed$coef

require(mvabund)
data(spider)
x<-apply(spider$x,2,rep,times=12)
dataspider<-data.frame(species=rep(names(spider$abund),each=28),abund=unlist(spider$abund),x)
out<-glmm(response.var="abund", grouping.var="species", fixed=~soil.dry+bare.sand, random=~1,
data=dataspider,distribution="poisson")

out2<-glmm(response.var="abund", grouping.var="species", fixed=~soil.dry+bare.sand,
data=dataspider,distribution="poisson")
fit<-manyglm(mvabund(spider$abund)~soil.dry+bare.sand,data=data.frame(spider$x),family="poisson")

helske/StateSpaceGLMM documentation built on May 17, 2019, 3:39 p.m.