logLik.emfrail: Log-likelihood for 'emfrail' fitted models

In tbalan/frailtyEM: Fitting Frailty Models with the EM Algorithm

Description

Log-likelihood for emfrail fitted models

Usage

## S3 method for class 'emfrail'
logLik(object, ...)

Arguments

object	An emfrail object
...	Other arguments

Details

The formula for the likelihood can be found in the manual which accompanies the package. Note that a constant is added. If we denote \bar{n} the total number of events and \bar{n_i} the total number of events at time point i, for each time point where events are observed, then this is equal to

\bar{n} - ∑_i \bar{n_i} \log \bar{n_i}.

This is mostly because of compatibility, i.e. to match the log-likelihood given by the survival package.

The df attribute of this object is equal to the number of regression coefficents plus 1. In general, the number of degrees of freedom for a frailty model is an unclear concept. For the coxph frailty fits, and in general for the shared frailty models fitted by penalized likelihood, the degrees of freedom is a number that depends on the penalization. However, even in that case, there is no straight forward interpretation or use of this quantity. The decision made here is because this would keep the likelihood ratio test for a covariate effect valid.

Value

An object of class logLik containing the marginal log-likelihood of the fitted model

tbalan/frailtyEM documentation built on Sept. 22, 2019, 5:57 a.m.