Causal Analysis and Ecological Fallacy in Cross-national Epidemiological Research

Abstract

This paper has two interrelated purposes. First, we examine the usefulness of a statistical technique derived from a causal model, i.e., path analysis, in the analysis of cross-national epidemiological data. With this method, the data analysis can be conducted according to the causal relations among the variables specified by the theory. In this way, the statistical analysis of the data and the testing of causal hypotheses are tied together more closely than when using other statistical techniques. Second, we scrutinize the importance of the analytical unit in the interpretation of the results, by comparing within-area results, based on individuals, with results calculated across the areas, using aggregate data. Because of the “ecological fallacy”, interpretations on individuals cannot usually be made on the basis of area level results. The results are demonstrated by using data from a cross-national study incorporating 48 subpopulations in seven countries.