Tuesday, February 21
Mathematics Colloquium
03:35 PM - 04:35 PM
Location: Science Center Room 323
Cost: Free

Speaker: Yassir Rabhi, University of Sherbrooke

Title: A semiparametric regression under biased sampling and random censoring: a local pseudo-likelihood approach

Abstract: Selection bias is often encountered in cross-sectional surveys and prevalent-cohort studies on disease duration. In regression, ignoring biased sampling leads to bias in estimating the lifetime and the regression function model of a population. Most of the studies were focused on modeling the regression either with completely known or completely unknown truncation distribution F_t. In this presentation, we introduce a new semiparametric regression estimator that combines the advantages of both approaches: retains the efficiency of the first and maintains some robustness to the form of F_t. The proposed approach models right-censored left-truncated data when F_t belongs to a known parametric class F_a with unknown parameter (s) a. We devise an estimator for the regression function based on a local pseudo-likelihood approach, and derive conditional MLE for a. Also, we develop a pseudo-likelihood estimator for the regression function when the sampling is subject to uniform truncations and random censoring. This estimator is shown to be more efficient than classical estimator for unspecified F_t. The asymptotic properties of the estimators are established. In addition, we provide a bandwidth selection method to choose the smoothing parameter. Several examples are discussed and assessed, on the basis of which we conclude that our semiparametric estimator, for F_t \in F_a, outperforms classical estimator for unspecified F_t. The proposed method is then applied to analyze a set of data on survival with dementia.

Refreshments are available at 3:00 PM in SC 313F.

The department colloquia are held every Thursday (excluding holidays) at 3:35 pm in room SC 323 unless otherwise noted. All are invited to attend. 

Contact Information:
Name: Paul Eloe