In the health and social sciences, “ecological inference” is the term used to refer to analyses that aim to make inference on the relationship between individual-level quantities using aggregate (group level) data. Such ecological inference is often subject to bias and imprecision, due to the lack of individual-level information in the data. These problems can be reduced by supplementing the aggregate-level data with small samples of data from individuals within the areas, which directly link exposures and outcomes.
In this talk, I will outline a new class of models – termed hierarchical related regressions (HRR) – for estimating individual-level associations using a combination of aggregate and individual data. The HRR models combine features of standard ecological regression models for aggregate data and multilevel models for clustered individual-level data, and have been shown to reduce bias and improve precision in many situations.
Two case studies will be discussed:
1. An investigation of the individual and contextual effects of socioeconomic factors on risk of chronic diseases such as cardiovascular illness and self-reported limiting long term illness, using a combination of individual-level data from the Health Survey for England, and aggregate data from the Census and Hospital Episode Statistics.
2. An investigation of the extent to which there was a realignment of Muslim voters away from Labour between the 2001 and 2005 British General Elections, using individual-level British Election Survey data with aggregate census data and constituency election results.
Presented by:
Nicky Best, Imperial College London
Date & time:
May 4, 2010 3:00 pm - May 4, 2010 4:30 pm
External seminars home