Dynamic Prediction for Discrete-Time Recurrent Event Data: Longitudinal Incidence and Prevalence of Infant Diarrhoea
Robin Henderson, University of Newcastle
Venue: Room A54, Postgraduate Statistics Centre, Lancaster University
Date: 26-02-2015, 4 - 5pm
This talk is based around the analysis of data on incidence and prevalence of infant diarrhoea, which is one of the main causes of
infant mortality in developing countries. The data were collected in three waves as part of an investigation into the effect of a sanitation
programme in Salvador, Brazil. We consider methods for the analysis when interest is mainly in prediction. The Aalen additive model provides an extremely simple and effective method for the determination of covariate effects for this type of data, especially in the presence of time-varying effects and time varying covariates, including dynamic summaries of prior event history. The method is weakened for predictive purposes by the presence of negative estimates, an obvious alternative being dynamic logistic regression, perhaps with a penalty to provide stability. We describe several measures of predictive ability for recurrent event data and
investigate the predictive capability of several different modelling approaches.