Ecob, R. A multilevel modelling approach to examining effects of area of residence on health and functioning. J. Roy. Stat. Soc. Series A. Vol 159, 1, 61-76.

A number (6) of measures of health and function from the baseline wave, in 1987 of the West of Scotland Twenty-07 study are examined. Multilevel models allow for the population structure and allow further examination of the variance at the postcode sector level (the primary sampling unit) after allowing for a range of individual and household characteristics. Area deprivation is included as an explanatory variable in some models. The variance at both the individual and area levels is allowed to vary both by age and gender. This analysis was made possible by recent developments in the software implementation of Multilevel Models using the ML3 software. Two (waist-hip ratio and reaction time) out of the six measures of health and function are shown to relate to area of residence (statistically significant variance component at the area level), this variation varying in turn according to age and gender. Four measures (waist-hip ratio, reaction time, self rated health and height) are related to area deprivation. Limiting longstanding illness and standardised respiratory function are related to neither area or area deprivation given individual characteristics.

Ecob, R Contribution to the discussion of Goldstein, H and Speigelhalter, D.J 'League tables and their limitations: statistical issues in comparisons of institutional performance' J Roy. Stat. Soc. Vol 159,3. 385-444.1994

This discussion contribution suggests an interpretation for the observed effect of primary school on between secondary school differences in exam results in terms of possibly unmeasured effects of catchment area.