Seminar
Nicholas Horton
"Principled data combination of multiple source reports
using manifest and latent variable regression models"
The talk will review regression-based methods for analyzing
multiple-source data. The term multiple-source data is used to
encompass all cases where data are simultaneously obtained from
multiple informants, or raters (e.g., self-reports, family members,
health care providers, administrators) or via different/parallel
instruments, indicators or methods (e.g., symptom rating scales,
standardized diagnostic interviews, or clinical diagnoses). This
is an important problem in many social science and medical research
areas, particularly health services research. Manifest regression
models for analyzing multiple source risk factors, special cases
of generalized linear models, albeit with correlated outcomes. In
addition, a series of latent variable models for multiple source
predictors using maximum likelihood will be proposed. The methods
are illustrated using datasets from psychiatric epidemiology and
environmental health research.
