Online Program Home
  My Program

All Times EDT

Abstract Details

Activity Number: 505 - A Variety of Problems in Statistical Inference
Type: Contributed
Date/Time: Thursday, August 6, 2020 : 10:00 AM to 2:00 PM
Sponsor: Biometrics Section
Abstract #313214
Title: Regression of Observations Below the Limit of Detection Through a Semiparametric Pseudo-Value Approach
Author(s): Sandipan Dutta* and Susan Halabi
Companies: Old Dominion University and Duke University
Keywords: Censored Regression; Limit of Detection; Pseudo-value; Tobit Regression; Serum Androgen
Abstract:

Before biomarkers can be used, the laboratory assays that measure their levels must go through development and analytical validation. One of the most critical performance metrics for validation of any assay is related to the minimum amount of values that can be detected and these are known as the limit of detection (LOD). A common approach for analyzing outcomes below the LOD is to use a parametric censored regression model, such as the Tobit regression model. Such parametric models heavily depend on the distributional assumptions and can result in loss of precision under model misspecifications. As a motivating real-life example from a prostate cancer clinical trial, we show that how an important relationship between serum androgen and a prognostic factor of overall survival is completely missed by the parametric model. We utilize a semiparametric pseudo-value based regression approach that effectively captures the important relationship between the serum androgen and the prognostic factor. Simulation results show that our method performs better than the parametric censored regression method and establishes itself as a powerful alternative to the standard parametric approaches.


Authors who are presenting talks have a * after their name.

Back to the full JSM 2020 program