A0620
Title: Extending age-specific logistic regression with complex event times
Authors: Yan Yuan - University of Alberta (Canada) [presenting]
Abstract: The motivation is the project of the Childhood Cancer Survivor Study, which aims at predicting primary ovarian insufficiency (POI) risk among female survivors. The unconventional right-censored age of POI leads to an age-varying risk set. Two extensions of the age-specific logistic regression are proposed, presented in a recent study, and the POI study data is analyzed by the proposed approaches. The efficiency and robustness of the two approaches were examined numerically and theoretically. The findings are verified using the POI study. The survival random forest (SRF) is used to obtain learning weight in the approaches. Simulations are conducted to compare SRF with other conventional methods for obtaining weights. Additionally, using kernel smoothing, smoothed age-varying effect size estimates are obtained for risk factors. The proposed approaches can be incorporated into the current Shiny app for predicting POI risk.
