A0365
Title: Efficient estimation for functional accelerated failure time models
Authors: Changyu Liu - The Hong Kong Polytechnic University (Hong Kong)
Xingqiu Zhao - The Hong Kong Polytechnic University (Hong Kong)
Guosheng Yin - The University of Hong Kong (Hong Kong)
Wen Su - City University of Hong Kong (Hong Kong)
Kin Yat Liu - The Chinese University of Hong Kong (Hong Kong) [presenting]
Abstract: A functional accelerated failure time model is introduced to analyze the impact of both functional and scalar covariates on the time until an event occurs. Regularity conditions are also established to ensure the model is identifiable. For efficient parameter estimation, a sieve maximum likelihood method that combines parametric and nonparametric coefficients with an unknown baseline hazard function within the likelihood function is created. This combination not only leads to significant numerical challenges but also introduces new theoretical difficulties. By formulating a comprehensive theoretical framework, these issues related to the bundled parameters are addressed, and the convergence rate of the proposed estimator is determined. Moreover, it is demonstrated that the finite-dimensional estimator is root-n consistent, asymptotically normal, and reaches the semiparametric information bound. The nonparametric optimality of the functional estimator is further shown, and an asymptotic simultaneous confidence band is constructed. The effectiveness of the inference procedures is assessed through extensive simulations and illustrated with a case study using data from the National Health and Nutrition Examination Survey.
