QIC and quasi-Likelihood for GEE
Calculate quasi-likelihood under the independence model criterion (QIC) for Generalized Estimating Equations.
QIC(object, ..., typeR = FALSE) QICu(object, ..., typeR = FALSE) quasiLik(object, ...)
object |
a fitted model object of class |
... |
for QIC and QICu, optionally more fitted model objects. |
typeR |
logical, whether to calculate QIC(R). QIC(R) is
based on quasi-likelihood of a working correlation R model. Defaults
to |
If just one object is provided, returns a numeric value with the
corresponding QIC; if more than one object are provided, returns a
data.frame
with rows corresponding to the objects and one column
representing QIC or QICu.
This implementation is based partly on (revised) code from packages yags (R-Forge) and ape.
Kamil Bartoń
Pan W. (2001) Akaike's Information Criterion in Generalized Estimating Equations. Biometrics 57: 120-125
Hardin J. W., Hilbe, J. M. (2003) Generalized Estimating Equations. Chapman & Hall/CRC
Methods exist for
gee
(package gee),
geeglm
(geepack),
geem
(geeM),
wgee
(wgeesel, the package's QIC.gee
function is used),
and
yags
(yags on R-Forge).
There is also a QIC
function in packages MESS and geepack,
returning some extra information (such as CIC and QICc). yags
and
compar.gee
from package ape both provide QIC values.
data(ohio) fm1 <- geeglm(resp ~ age * smoke, id = id, data = ohio, family = binomial, corstr = "exchangeable", scale.fix = TRUE) fm2 <- update(fm1, corstr = "ar1") fm3 <- update(fm1, corstr = "unstructured") # QIC function is also defined in 'geepack' but is returns a vector[6], so # cannot be used as 'rank'. Either use `MuMIn::QIC` syntax or make a wrapper # around `geepack::QIC` QIC <- MuMIn::QIC ## Not run: QIC <- function(x) geepack::QIC(x)[1] ## End(Not run) model.sel(fm1, fm2, fm3, rank = QIC) ## Not run: # same result: dredge(fm1, m.lim = c(3, NA), rank = QIC, varying = list( corstr = list("exchangeable", "unstructured", "ar1") )) ## End(Not run)
Please choose more modern alternatives, such as Google Chrome or Mozilla Firefox.