A1221
Title: A quantile cure model with fully or partially functional covariate effects
Authors: Chyong-Mei Chen - Institute of Public Health (Taiwan) [presenting]
Abstract: The quantile regression has several attractive features, such as its ability to allow covariate effects to vary at different quantile levels and to handle heteroscedasticity in data easily, which make it a viable alternative when analyzing data with continuous outcomes in recent years. In particular, it has been used in modeling survival data with and without a cured fraction. Novel estimating equation approaches are proposed to estimate a mixture cure model where the latency survival time is modeled by a quantile regression. The proposed estimation methods enjoy a double robustness in the sense that a misspecification in one of the two parts in the mixture cure model will not affect the estimation in the other part. The methods do not require the global log-linear assumption in the quantile regression, and they allow mixed effects of functional and constant effects in the regression when the log-linear assumption is held at an interval of quantile levels. The asymptotic properties of the proposed estimators are established. The simulation studies demonstrated the double robustness and the efficiency gains in the proposed estimators. An application of the proposed model and methods to data from a lung cancer study revealed new and interesting findings that were not reported in a previous analysis of the data.
