SEMINAR: The University of Reading,
School of Applied Statistics, room AS-G03
Thursday, June 3, 2pm
Nicky Best (joint work with Chris Jackson and Sylvia Richardson)
Department of Epidemiology and Public Health
Imperial College London.
http://www1.imperial.ac.uk/medicine/people/n.best.html
Improving ecological inference using individual-level data
Abstract:
Estimating the relationship between individual level quantities
using aggregate-level data, known as ecological inference, is
often subject to bias. Such bias can be reduced by supplementing
the aggregate-level data with a sample of data from individuals
that directly links exposures and outcomes. This talk will present
a Bayesian hierarchical model framework to deal with this situation,
and will extend recent work by Wakefield (2004) to consider the
joint effects of binary and continuous exposures in ecological
inference. Results of a simulation study will be presented,
using a variety of realistic conditions to determine when the
aggregate data are sufficient for accurate inference, and
individual-level information is also requited. The method
will be illustrated by combining aggregate data from the 1991
UK census with individual level data from the Health Survey
for England on limiting long-term illness, socio-economic
deprivation and ethnicity.
References:
Wakefield JC. (2004). Ecological inference for 2 x 2 tables.
Journal of the Royal Statistical Society, Series A (to appear).
Jackson CH, Best NG and Richardson S. (2004). Improving ecological
inference using individual-level data. Submitted.
All welcome.
Travel information: http://www.rdg.ac.uk/Statistics/info/direc.html
Other seminars: http://www.rdg.ac.uk/Statistics/diary/seminars.html
Posted by: Howard Grubb ([log in to unmask])
|