coxph and
survReg models.
frailty.gamma(x, sparse=(ngroup>5), theta, df, eps=1e-05,
method=c("em", "aic", "df", "fixed"), ...)
"fixed" method requires that
theta be given explicitly (and is assumed
when
theta is specified), the
"df" method is used when degrees of freedom
is explicitly set.
The
"em" method computes the maximum likelihood estimate, and achieves
the same solution as the EM algorithm found in several references.
This is an iterative method that searches for the maximum of
a profile likelihood.
The
"aic" method chooses the variance (and hence the degrees of freedom) based
on Akaike's information criteria.
trace=T is the most
usual.
(The set of available arguments depends on the specific control function).
For the AIC method the "caic=T" argument may be used to choose the corrected
AIC criteria.
"coxph.penal".
When using frailty terms with a
survReg model,
only
method="aic" should be used.
The
method argument can be set
in the call to
frailty in the formula.
The
method argument will be passed down to the specific
frailty function via the
... argument.
fit1 <- coxph(Surv(time, status) ~ age + sex + frailty(inst, df=4),
data=lung, na.action=na.exclude)
round(fit1$frail,2)