Semiparametric cure model

semicure

Description

Fit a semiparametric cure model.

Usage

semicure(formula = formula(data), cureform, data = sys.parent(), link = "logit", model = c("glm", "gam"), emmax = 100, eps = 1e-4, debug = T, tail = c("zero", "none"), savedata = T, ...)

Arguments

formula: a formula expression as for other regression models. It specifies the effects of covariates (on the righthand side of the ~) on the failure time (on the lefthand side of the ~) of uncured patients through the proportional hazards assumption.

cureform: a formula expression as for formula without the response. It specifies the effect of covariates (on the righthand side of the ~) on the cure rate. A covariate may be used in both formula and cureform. The absence of cureform implies no cured patients. The default assumes no cured patient.

data: a data frame in which to interpret the variables named in the formulas.

link: the link function relates the cure rate to covariates in cureform. The default is the logit link.

model: the model to fit the effects of cobariates on the cure rate. It can be one of "glm" and "gam". The default is "glm". If there are additive effects of covariates in cureform, "gam" should be used.

emmax: the number of iterations in the EM algorithm.

eps: tolerance for convergence. Iteration stops once the relative change in deviance is less than eps.

tail: the method to complete the tail of the baseline survival function. It can be one of "zero" and "none". The default is "zero".

debug: a logical variable. If TRUE, some debug information printed.

savedata: a logical variable. If TRUE, the orginal data set combined with the estimated cure rate for each patient is saved into the final semicure object.

Value

an object of class semicure is returned. It can be examined by coef(), print(), summary() and predict().

References

Peng, Y. (2003), Fitting semiparametric cure models, Computational Statistics and Data Analysis 41:481--490.

Examples

> semicure(Surv(time, cens) ~ transplant, ~ transplant, data = goldman.data) 
Program is running ............. done. 
Call: 
semicure(formula = Surv(time, cens) ~ transplant, cureform = ~ transplant, 
        data = goldman.data) 
Failure time distribution model: 
Call: coxph(formula = Surv(time, cens) ~ transplant + offset(log(uncureprob)), 
        subset = uncureprob != 0, method = "breslow") 

                 coef    exp(coef)  se(coef)     z     p 
transplant        0.317    1.37        0.13       2.43  0.015 
Likelihood ratio test=5.96 on 1 df, p=0.0146 n= 80 
Cure probability model: 
Call: 
glm(formula = uncureprob ~ transplant, family = binomial(link = "logit")) 
Coefficients: (Intercept) transplant 
                1.216835   0.1884882 

Degrees of Freedom: 91 Total; 89 Residual 
Residual Deviance: 91.62084

Semiparametric Cure Model