Abstract:
|
In the analysis of clustered time-to-event, it is common practice to postulate a random effect model to handle inherent inter-cluster correlation. Specification and interpretation of such a conditional model, however, is dependent on the distribution of random effects which though often overlooked, is difficult to verify. Besides, one may also be interested in making inference on the marginal covariate effect in addition to subject specific inference. In this paper, we formulate a joint logistic-normal and a combined frailty model to analyse interval censored data with right truncation. We further extend our work to the Marginalized Multilevel Models (Heagergty and Zegger , 2000) and Bridge distribution for random effects (Huang and Louis, 2003) to present a unified approach that allows parameter estimates to be interpreted both at the marginal and conditional levels. The methods are applied to the analysis of data from the HET-CAM$^{VT}$ experiment which examines the viability of using a fertilized chicken egg to predict irritation potential of a given compound formulation. Of interest is the times to occurrence of various irritation indicators. Besides interv
|
ASA Meetings Department
732 North Washington Street, Alexandria, VA 22314
(703) 684-1221 • meetings@amstat.org
Copyright © American Statistical Association.