parfm (version 2.7.7)

select.parfm: AIC and BIC values of several Parametric Frailty Models

Description

The function select.parfm() computes the AIC and BIC values of parametric frailty models with different baseline hazards and different frailty distributions.

Usage

select.parfm(formula, cluster=NULL, strata=NULL, data, inip=NULL, iniFpar=NULL,
             dist=c("exponential", "weibull", "inweibull", "frechet", "gompertz", 
                    "loglogistic", "lognormal", "logskewnormal"),
             frailty=c("none", "gamma", "ingau", "possta", "lognormal"),
             method="BFGS", maxit=500, Fparscale=1, correct=0)

Value

An object of class select.parfm.

Arguments

formula

A formula object, with the response on the left of a ~ operator, and the terms on the right. The response must be a survival object as returned by the Surv function.

cluster

The name of a cluster variable in data.

strata

The name of a strata variable in data.

data

A data.frame in which to interpret the variables named in the formula.

inip

The vector of initial values. First components are for the baseline hazard parameters according to the order given in 'details'; Other components are for the regression parameters according to the order given in 'formula'.

iniFpar

The initial value of the frailty parameter.

dist

The vector of baseline hazards' names. It can include any of weibull, inweibull (alias frechet), exponential, gompertz, loglogistic or lognormal.

frailty

The vector of frailty distributions' names. It can include any of: none, gamma, ingau, possta or lognormal.

method

The optimisation method from the function optim().

maxit

Maximum number of iterations (see optim()).

Fparscale

the scaling value for the frailty parameter in optim(). Optimisation is performed on Fpar/Fparscale.

correct

A correction factor that does not change the marginal log-likelihood except for an additive constant given by #clusters * correct * log(10). It may be useful in order to get finite log-likelihood values in case of many events per cluster with Positive Stable frailties. Note that the value of the log-likelihood in the output is the re-adjusted value.

Author

tools:::Rd_package_author("parfm")

References

Munda M, Rotolo F, Legrand C (2012). parfm: Parametric Frailty Models in R. Journal of Statistical Software, 51(11), 1-20. DOI <tools:::Rd_expr_doi("10.18637/jss.v051.i11")>

See Also

parfm, ci.parfm, predict.parfm

Examples

# \donttest{
data(kidney)
kidney$sex <- kidney$sex - 1

models <- select.parfm(Surv(time,status) ~ sex + age, 
                       dist = c("exponential", 
                                "weibull",
                                "inweibull",
                                "loglogistic", 
                                "lognormal", 
                                "logskewnormal"),
                       frailty = c("gamma", 
                                   "ingau", 
                                   "possta", 
                                   "lognormal"),
                       cluster = "id", data = kidney)
models
plot(models)
# }