A0951
Title: Variable selection for the generalized odds rate non-mixture cure model with current status data
Authors: Xuewen Lu - University of Calgary (Canada) [presenting]
Saba Saghatchi - University of Calgary (Canada)
Jingjing Wu - University of Calgary (Canada)
Abstract: Current status or case I interval-censored data are common in labour economics, clinical trials, and hospital visits. In some cases, there exists a cured sub-population where individuals never experience the event of interest. Meanwhile, when the current status data is modeled, one may also encounter an extraordinarily large number of risk factors; variable selection is desired in the model building. The purpose is to study variable selection methods for the generalized odds rate non-mixture cure model with current status data using penalized semiparametric likelihood function when the dimension of the covariates is diverging. The proposed model encompasses the proportional hazards (PH) and proportional odds (PO) non-mixture cure models as special cases. The broken adaptive bridge (BAR) penalty is utilized for regularization, and the asymptotic properties of the resultant estimators of regression parameters, including the oracle and group properties, are studied. The sieve method based on Bernstein polynomials is employed to estimate the unknown cumulative distribution function. To facilitate the computation, a novel penalized expectation maximization (EM) algorithm is implemented. Furthermore, a simulation study is conducted to assess the finite sample performance of the proposed method and compare it with other existing penalized methods such as Lasso, ALasso, and SCAD. Finally, the method is applied to a real data set for illustration.