TIME-SERIES REGRESSION FOR COUNTS - AN INVESTIGATION INTO THE RELATIONSHIP BETWEEN SUDDEN-INFANT-DEATH-SYNDROME AND ENVIRONMENTAL-TEMPERATURE

Zeger's approach to the regression of count data which are serially ordered in time is discussed and compared with an empirical method. The methods are used to investigate the relationship between sudden infant death syndrome (SIDS) and environmental temperature in three data sets: for England and Wales 1979-83, England and Wales 1984-86 and New South Wales 1981-85. The results imply a strong negative relationship, in which the deaths due to SIDS increase 2-5 days after the environmental temperature goes down. The data are then explored to see whether there is a threshold in the temperature effect and to examine the consistency of the effect over time.