Relationship between sickness absence and meteorological factors.

fault in the methods of Wright and Wright and Boyd is the failure to eliminate broad seasonal trends (e.g., by using deviations from a moving average or some parametric trend) before investi- gating correlations between disease and meteorological factors. Spicer (1959) draws attention to the need to use such deviations, sometimes called residuals, and this approach has been adopted in the regression model described below. MORBIDrrY STUDIEs Holland, Spicer, and Wilson (1961) used two measures of the monthly occurrence of disease- admissions to the Emergency Bed Service of a London hospital and admissions to the sick quarters in three Royal Air Force stations. The climatic factors considered were average temperature, absolute humidity, rainfall, barometric pressure, sun, and pollution (smoke), and analysis of covari- ance was used to eliminate broad seasonal and 238 copyright.