Department of Statistics Seminar
North Carolina State University

presents

Dr. Cassandra Arroyo

Morehouse School of Medicine

"Optimal Design for Bivariate Longitudinal Data When Some of the Outcomes Will Be Missing by Design"

ABSTRACT

Designed missingness arises often in many different biomedical applications, mostly in the form of ancillary validation substudies. While such designed missingness has been well studied for predictors, relatively little has been done when the designed missingness occurs in the outcome. This issue has been discussed in the sociology literature under the heading of matrix sampling, but has been restricted to the single group context. We propose a latent variable model for analyzing bivariate longitudinal data when one of the outcomes is missing by design. We then use the method to motivate a discussion of optimal design for studies where one of the outcomes is subject to designed missingness and offer practical advice for investigators to consider when faced with this issue.

Friday, October, 14, 2005

3:35 - 4:35 pm

206 Cox Hall

Refreshments will be served on the second floor of Dabney Hall (left of Room 222) at 3:00 pm.