Department of Statistics Seminar
North Carolina State University

presents

Niels Keiding

University of Copenhagen

"Selection Effects and Nonproportional Hazards in Survival Models and Models for Repeated Events"

ABSTRACT

In survival analysis, deviations from proportional hazards may sometimes be explained by unaccounted random heterogeneity, or frailty. Proportional hazards at the individual level leads to converging hazards at the population level under random multiplicative heterogeneity with finite expectation. However, direct statistical inference is hampered by identifiability problems which are avoided with accelerated failure time modelling, the latter also yielding easily interpretable results. A 'standard' survival analysis example on malignant melanoma illustrates these points.

For models for repeated events the random heterogeneity would be assumed to be better determined and work is in progress, partly motivated by a recent study on repeated psychiatric admissions for manic-depressive patients. I shall describe that study, where the focus is whether the apparent increasing admission intensity with increasing number of admissions is a selection effect only or does have an interpretation at the individual level.

Tuesday, November 17, 1998

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.