We address a problem in inference from retrospective surveys where the value of a covariate is measured at the date of the survey but is used to explain events that have occurred long before the survey. This causes bias because the value of the current-date (anticipatory) covariate does not follow the temporal order of events. We propose a dynamic Bayesian modelling approach that allows effects of the anticipatory covariate to vary over time and, thereby, estimate their values at the time of the event of interest. The issues are illustrated with data on the effects of education attained by the survey-time on divorce risks among Swedish men. The overall results show that failure to adjust for the anticipatory nature of education leads to underestimation of relative risks of divorce across educational levels. The seminar will focus more on the models as the empirical results were presented in a previous seminar by the co-author.


Tid: 6 november, 2019, kl. 13-14 Plats: B705