Maximum Likelihood Estimation in Semiparametric Transformation Models for the Cumulative Incidence of Competing Risks

• Activity Number: 296
• Type: Contributed
• Date/Time: Tuesday, August 5, 2014 : 8:30 AM to 10:20 AM
• Sponsor: Section on Nonparametric Statistics
• Author(s): Lu Mao*+ and Danyu Lin
• Companies: University of North Carolina at Chapel Hill and University of North Carolina
• Keywords: Competing risks ; Cumulative incidence ; Nonparametric maximum likelihood ; Profile likelihood ; Semiparametric inference ; Transformation models

Abstract

In analyzing competing risks data, interest has been increasingly focused on the cumulative incidence of each failure type, instead of the cause-specific hazard. For semiparametric regression, the proportional hazards model of Fine and Gray (1999) has become a popular choice for data analysis. This approach, however, uses inverse probability censoring weighting (IPCW) and thus its validity relies on correct modeling of the censoring distribution. In addition, their estimators are not efficient. In this work, we propose a flexible class of semiparametric transformation models and derive efficient estimators through nonparametric maximum likelihood estimation (NPMLE). We propose a simple and efficient algorithm for computing the NPMLE via the profile likelihood method. We establish consistency, asymptotic normality and semiparametric efficiency of the estimators. Extensive simulations show that the procedure performs well in finite samples. Finally, a well-known bone marrow transplantation dataset is analyzed using the proposed method.