On the Use of Fractional Polynomial Models in Time-Series Studies of Particulate Matter and Mortality

• Activity Number: 581
• Type: Contributed
• Date/Time: Wednesday, August 1, 2012 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Ayano Takeuchi*+
• Keywords: Particulate Matter ; Fractional Polynomial ; Generalized Additive Models ; Accute Effect

Abstract

Last decade, it has been known that fine particulate matter (PM2.5) exposure is associated with daily counts of death from the result of various time series studies. Most studies use generalized linear model (GLIM) with natural spline or generalized additive model (GAM) with spline function for adjusting seasonal effects and weather conditions, because it is well known that they have great influence on health effect. Comparison of both models are often implemented, and most of the results suggests that GLIM overestimate the effect of PM2.5 (or GAM cause over fitting and under estimate the effect). In this study, we adopt fractional polynomial model (FPM), GAM and GLIM to apply daily time series study of PM2.5 data, and compare simulation results. The effect becomes larger in order of GAM, FPM, GAM, and GAM deeply dependent on the degree of smoothness. Time series data contain daily counts of death and daily concentrations of air pollutants in Japanese 20 cities from 2001 to 2007.