Finite mixture, zero-inflated Poisson and hurdle models with application to SIDS

  • Authors:

  • M. L. Dalrymple;I. L. Hudson;R. P. K. Ford

  • Affiliations:

  • Department of Mathematics and Statistics, University of Canterbury, Private Bag 4800, Christchurch, New Zealand;Department of Mathematics and Statistics, University of Canterbury, Private Bag 4800, Christchurch, New Zealand;Community Paediatric Unit, Christchurch Public Hospital, Christchurch, New Zealand

  • Venue:
  • Computational Statistics & Data Analysis
  • Year:
  • 2003

Quantified Score

Hi-index 0.03

Visualization

Abstract

This study examines the incidence of sudden infant death syndrome (SIDS) in Canterbury (1973-1989) in relation to climate. Three mixture models (Finite Mixture, Zero-inflated Poisson and Hurdle) are used as novel methods which are able to highlight differential effects of climatic covariates between months of SIDS and no SIDS. These methods accommodate the extra zeros, heterogeneity and autocorrelation found in the SIDS series. Mixture models are comprehensive methods applicable to many discrete chronological series including the Canterbury SIDS data. This analysis leads to a better understanding of the association between climate and SIDS deaths.Results show a deviance-temperature (a measure of extreme change from the fortnightly average) is significantly associated with SIDS risk (p 0.005). Months where there is a high deviance-temperature are associated with increased risk of SIDS, compared to months where the temperature has remained reasonably constant. This finding is consistent with the theory that hyperthermia, or overheating of infants leads to increased SIDS risk. In months where at least one SIDS death occurs, increased humidity leads to increased risk of SIDS (p 0.001).