It is well known that statistical analysis based on aggregated data, such as area or group means, may be invalid because of the ecological fallacy. This fallacy occurs when analyses based on area level means give conclusions very different from those that would be obtained from an analysis of unit level data, if they were available. Whilst the size of such effects has been investigated, little progress has been made in developing methods which can be applied to group level data to estimate unit level relationships. In this paper it is shown how the differences between unit and group level analyses may be explained through a model which incorporates the effect of variables which characterize to which areal units individuals belong, together with area level characteristics which produce correlations between individuals in the same areal unit or group. Methods are suggested which, together with some auxiliary information, can be used to adjust area level analyses to provide less biased estimates of unit level parameters. In addition, the identification of a small and convenient set of grouping variables is an important aspect of the proposed methodology and procedures are suggested to achieve this. These grouping variables may be important in providing a substantive explanation of the population structure. The result is a strategy for the analysis and adjustment of aggregation effects in multivariate statistical analysis.