Random Effect Term for a Survival Model

DESCRIPTION:

Fit a random effect term for a variable, where the random effect comes from the Gaussian distribution. This function is specific to the coxph and survReg models.

USAGE:

frailty.gaussian(x, sparse=(ngroup>5), theta, df, eps=1e-05,  
     method=c("reml", "aic", "df", "fixed"), ...)  

REQUIRED ARGUMENTS:

x
variable describing the groups. A separate indicator variable is fit for each group.

OPTIONAL ARGUMENTS:

sparse
use a sparse matrix method of solution. This parameter saves considerable time and memory when the number of groups is large, for example, a problem with one group per family and 200 families.
theta
variance of the random effect.
df
degrees of freedom for the random effect.
eps
convergence criteria for the outer loop of the algorithm.
method
method of choosing the variance. The "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 "reml" method uses McGilchrist's restricted maximum likelihood equation to choose a value of the variance. This is an iterative method that searches for the zero of an equation. The "aic" method chooses the variance (and hence the degrees of freedom) based on Akaike's information criteria.
...
optional arguments for the control function, of which 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.

VALUE:

an object with class "coxph.penal".

SIDE EFFECTS:

If sparse computation is chosen, then the coefficients for the sparse terms are not printed by the default print and summary methods.

NOTE:

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.

SEE ALSO:

, , .

EXAMPLES:

fit1 <- coxph(Surv(time, status) ~ age + sex + frailty(inst,  
       dist='gauss', df=4), data=lung, na.action=na.exclude) 
round(fit1$frail,3)