Gebrenegus Ghilagaber, PhD, Associate Professor, Stockholm University.

Bayesian Adjustment of Anticipatory Covariates in Analyzing Retrospective Data

Abstract:

In retrospective surveys, records on important variables like
respondent's educational level and social class refer to what is
achieved by the date of the survey. These variables are then used as
covariates in investigations of behaviour - such as marriage and
divorce - in life segments that have occurred before the
survey-date. However, such anticipatory (current-dated) covariates
become problematic because they don't follow the temporal order of
events. In the present work, we specify a continuous-time Markov model
for the incompletely observed (anticipatory) time-varying covariates
and implement standard Bayesian data augmentation techniques. The
issues are illustrated by estimating effects of anticipatory
educational level on divorce-risks within the framework of a
multiplicative (log-linear) piece-wise-constant intensity
model. Results show that ignoring the time-inconsistency of
anticipatory covariates may seriously plague the analyses because the
relative risks across the anticipatory educational level are
overestimated.

Joint work with Johan Koskinen (PhD), Stockholm University &
University of Melbourne.