DEPARTMENT OF MEDICAL STATISTICS
DE MONTFORT UNIVERSITY, LEICESTER
Research Students' Seminar Announcement
Time: Wednesday, 5th May 1999, 1 p.m.
Place: James Went Building, DMU Leicester
Room JW 6-06
Speaker: Professor Mark Espeland
Wake Forest University, USA
Title: Statistical and design issues associated
with multivariate outcome measures:
Analysis of carotid ultrasonography data
Abstract
Ultrasonography of the carotid arteries provides non-invasive images
of the extent of local atherosclerosis. Many epidemiologic cohort
studies and clinical trials have developed outcome measures based on
this methodology. However, difficult issues arise in identifying
repeatable and valid empirical outcomes from images. Measurements are
inherently multivariate and subject to missing data mechanisms that
may not be ignorable. Hierarchical linear models are most
appropriate; however, many assumptions must be made concerning
cross-sectional and longitudinal covariance structures. Using data
from the Asymptomatic Carotid Artery Progression Study (ACAPS) trial
and the Carotid Artery Follow-up Study (CAFUS), I will discuss several
approaches that have been developed to address these issues. Such
data serve as interesting test cases for examining more general
analytical issues and underscore that care in addressing multivariate
incomplete data can substantially improve the statistical power and
statistical efficiency of clinical trials and cohort studies.
Info: Nick Longford, [log in to unmask]
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|