mixcure.penal.2d.nested.lrt: mixcure.penal.2d.nested.lrt

In ChangchangXu-LTRI/Mixcure: Maximum (penalized) likelihood estimation of Weibull mixture cure model

mixcure.penal.2d.nested.lrt

Description

Likelihood ratio test (LRT) or penalized likelihood ratio test (PLRT) for estimated parameters, that outputs p-values based on asymptotic chi square distribution with df=2. Under nested deviance method, to perform LRT for two parameters alpha and beta of a MC model with information martix of dim(I)=k, for MLE the reduced model assumes alpha=beta=0 and dim(I)=k-2; for FT-PLE the reduced model assumes alpha=beta=0 and dim(I)=k for the penalty term.

Usage

mixcure.penal.2d.nested.lrt(formula, data, init, pl, loglik, iterlim = 200)

Arguments

formula	MC model specification in the form of a conventional survival model, i.e., Surv()~. Note the whatever x variables specified in the latency part of the model are assumed to be also regressed on in the incidence part, and vice versa.
data	a data object in the form of dataframe.
init	a vector of initial values input for the optimization of the vector of parameters specified in the likelihood. Usually, specified in the order of 1) regression parameters for the incidence part, 2) regression parameters for the latency part, and 3) the shape parameter of Weibull hazard function for the latency part.
pl	if TRUE, uses FT-PL; otherwise uses the usual likelihood for the model.
iterlim	specifies the number of maximum iterations for model parameter optimization. DEFAULT set to 200.
loglik	specifies the (penalized) loglikelihood value of the full model or non-restrected model.

Details

see mixcure.penal.est

Value

The mixcure.penal.1d.nested.lrt function returns an object of class 'mixcure.1d.nested.lrt' that encompasses a list of the followings:

coefficients

returns a table, that contains the (penalized) loglikelihood value of the reduced model, the (penalized) loglikelihood ratio in compared to the (penalized) loglikelihood value of the full model and the p-value under LRT or PLRT for the two parameters associated with every variable.

Author(s)

Changchang Xu

References

Heinze(2020), Firth(1993)

Examples

# Begin example

data(ANNbcBMdat1)
mc.mle1 <- mixcure.penal.est(Surv(Time, CENS == 1) ~ Her2,data=ANNbcBMdat1,init=c(1,-0.1,-10,1,1), pl=F)
mc.ple1 <- mixcure.penal.est(Surv(Time, CENS == 1) ~ Her2,data=ANNbcBMdat1,init=c(1,-0.1,-10,1,1), pl=T)

# Perform LRT for the MC model using maximum likelihoods:
mc.mle.2dlrt1 <- mixcure.penal.2d.nested.lrt(Surv(Time, CENS == 1) ~ Her2,data=ANNbcBMdat1, mc.mle1$loglikelihood, init=c(1,-0.1,-10,1,1), pl=F)

# Now for the bias-reduced model using penalized maximum likelihood:
mc.ple.2dlrt1 <- mixcure.penal.2d.nested.lrt(Surv(Time, CENS == 1) ~ Her2,data=ANNbcBMdat1, mc.ple1$loglikelihood,init=c(1,-0.1,-10,1,1), pl=T)

ChangchangXu-LTRI/Mixcure documentation built on April 22, 2022, 3:33 p.m.