A0254
Title: Variable selection under proportional odds model with right-censored survival data
Authors: Ming-Yueh Huang - Academia Sinica (Taiwan) [presenting]
Abstract: The focus is on the proportional odds model for right-censored survival data. This model directly estimates survival probabilities and is particularly useful when the assumptions underlying the widely used Cox proportional hazards model are violated. Rather than using the survival responses directly, the induced counting processes are considered, which follow a logistic regression model at each time point and yield a simpler likelihood function for estimation. This approach allows for the direct application of existing penalization methods such as LASSO, Elastic-Net, adaptive LASSO, and adaptive Elastic-Net to select relevant covariates without computing complex likelihood functions. To handle incomplete right-censored data, the imputation method is adopted for induced counting processes proposed by a past study. After imputation, both the coefficients and the relevant covariates are simultaneously estimated using a straightforward penalized weighted least squares criterion. Overall, the proposed method significantly reduces the computational burden compared to existing estimation and variable selection procedures for the proportional odds model with right-censored survival data.
