Time series modelling of disease transmission

Bärbel Finkenstädt

University of Warwick

A general framework for identifying and estimating models for disease incidence time series is presented. In the first part of the talk we focus on a stochastic model to explain the spatio temporal spread of childhood diseases. To formulate a realistic model for infectious diseases one must specify the form of transmission both within and between communities. Using the measles time series from cities in England and Wales we show that a simple coupling procedure can account for much of the spatial interaction. This work is joint with Alexander Morton (University of Warwick) and Bryan Grenfell (University of Cambridge).

In the second part of the talk we focus on the prediction of Influenza epidemics which is made substantially more difficult by antigenic drift and shift. We have developed a pair of simple stochastic Susceptible--Infected--Recovered—Susceptible (SIR-S) epidemic models to address this problem. The model incorporates constant loss of immunity due to antigenic drift but can be extended to allow for time-dependent shifts in the level of immunity. An ansatz for maximum likelihood estimation is presented and the model parameters are estimated from weekly surveillance data from four different European countries. This work is joint with Alexander Morton and David Rand (University of Warwick).


Go back to Statistics seminars

Go to Department of Statistics home page