Online Program Home
  My Program

All Times EDT

Abstract Details

Activity Number: 337 - Approaches for Modeling Clustered and Longitudinal Data
Type: Contributed
Date/Time: Wednesday, August 5, 2020 : 10:00 AM to 2:00 PM
Sponsor: Biometrics Section
Abstract #312836
Title: Joint Models for Multiple Ratings of a Discrete Diagnostic Test and Associated Auxiliary Variables
Author(s): Xianling Wang* and Gong Tang
Companies: University of Pittsburgh and University of Pittsburgh
Keywords: Discrete diagnostic tests; Latent class models; Auxiliary information

Discrete diagnostic tests without a gold standard such as tumor grading are important prognostic factors, but suffer from intra-rater and inter-rater reproducibility. With multiple ratings, the underlying class prevalence and raters’ classification probabilities are estimable up to a permutation of the underlying truth via latent class models (Kruskal, 1977; Dawid, 1979). When an auxiliary variable associated with the underlying classes is also observed, we propose a joint model to achieve global identification of the parameter estimates and improved efficiency. Remedy to a specific violation of the conditional independence assumption on the independent raters is also provided. The methods are illustrated via analysis of a tumor grade reading data from a study of the National Surgical Adjuvant Breast and Bowel Project (NSABP). The improved efficiency in parameter estimates is also demonstrated in simulation studies.

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

Back to the full JSM 2020 program