Ecological analysis involves analysing aggregate datafor groups of individuals to make inferences about relationshipsat the individual level. Often the results of such analyses givebadly biased estimates. This paper will consider the sources ofbias in linear regression analysis using aggregate data. The roleof variation of the individual level relationships between groupsand the consequent within-group correlations and how these arerelated to auxiliary variables that characterise the differencesbetween groups is considered. A method of adjusting ecologicalregression for the effects of auxiliary variables is described andevaluated using data from the 1991 Australian Census.
展开▼