A comparison of existing measles models

Loading...
Thumbnail Image
Date
2017
Authors
Allotey, Clifford
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
With improvements in computational power, more and more mechanistic models are being formulated to explain physical and biological phenomena. In this thesis, we assess the strengths and weaknesses of different well-known models of pre-vaccine era measles dynamics. We investigate the assumptions inherent in each model, how well they fit to pre-vaccine era data and how well simple extensions of them perform when extended to modeling vaccine era dynamics. The four focus models we studied were (1) the standard deterministic SEIR models with school-term forcing, (2) the stochastic SEIR model with school-term forcing, (3) the Bjornstad et al. [2002] time series SIR model with a general time-varying transmission rate and (4) the He et al. [2010] model which incorporates school-term forcing, cohort effect and infection from outside the population into the model. We found that the He et al. [2010] model, fitted using maximum likelihood estimation, is the best model in terms of likelihood and the Akaike information criterion. However, this model was also the lowest ranked among all four models when comparing fit using residuals, leading to some open questions on possible trade offs between noise and likelihood.
Description
Keywords
Epidemiology, Mathematics, Measles models
Citation