M J Campbell1, A Tobías. 1. University of Sheffield, UK. m.j.campbell@sheffield.ac.uk
Abstract
BACKGROUND: The relationship between air pollution and health is usually analysed by time series analysis. However, temporality (i.e. looking to see whether the cause precedes the effect) is rarely examined in detail. METHODS: Models were fitted to daily levels of black smoke, nitrogen dioxide, sulphur dioxide, ozone and mortality from Barcelona 1986-1995 to account for seasonality, environmental temperature, days of the week and influenza epidemics. Cross-correlations of the residuals were plotted for different lags. RESULTS: Clear evidence of a temporal relationship between mortality and air pollution found for all four pollutants in that changes in the pollutant preceded changes in mortality, implying causality. However the pattern of dependence was different for each pollutant. CONCLUSION: The cross-correlation plot is a useful tool in the analysis of air pollution time series.
BACKGROUND: The relationship between air pollution and health is usually analysed by time series analysis. However, temporality (i.e. looking to see whether the cause precedes the effect) is rarely examined in detail. METHODS: Models were fitted to daily levels of black smoke, nitrogen dioxide, sulphur dioxide, ozone and mortality from Barcelona 1986-1995 to account for seasonality, environmental temperature, days of the week and influenza epidemics. Cross-correlations of the residuals were plotted for different lags. RESULTS: Clear evidence of a temporal relationship between mortality and air pollution found for all four pollutants in that changes in the pollutant preceded changes in mortality, implying causality. However the pattern of dependence was different for each pollutant. CONCLUSION: The cross-correlation plot is a useful tool in the analysis of air pollution time series.