Peak flow rates of school children and environmental factors : mixed-effects modelling based on clusters of environmental variables

الملخص EN

The purpose of this study is to demonstrate a new approach That uses model-based clustering of environmental variables instead of the variables themselves (as a method of data reduction), and to develop a multi-level model based on these clusters in order to investigate the relationship between groups of environmental variables and peak flow rate for school children.

The approach is applied to the same data in ward et al (2002) in which the peak flow rates were twice daily (am and pm) recorded for a period of 56 consecutive days of a cohort of 150, 9 years old, children was selected at five schools in Birmingham and sand well.

Our approach involves a modeling strategy carried on in two major stages, namely; model-based clustering based on expectation maximization (em) algorithm to reduce the huge number of variables, and residual maximum likelihood is then applied to a mixed-effects multilevel model.

Stage one reduced the 42 independent variables and their lags to 5 main groups.

In stage two, investigated the relations between these groups and the peak flow as a dependent variable.

Our results reveal the followings: the higher the pollution the lower the peak flow, one specific school has shown a common low peak flow among the children, and we find as expected significant difference between morning and evening peak flow.