An alternative to ecologic regression analysis of mortality rates

A number of recent papers have used geographically defined data and linear models to study the relationship between a series of epidemiologic factors and the frequency of disease. This "ecologic regression" approach involves serious problems of interpretation. An alternate approach is discussed that does not depend on statistical models, produces easily interpreted results and yields statistical summaries that approximately parallel regression analysis. This alternative procedure is illustrated by using acute lymphocytic leukemia mortality data on a county level from among white females less than five years of age between 1969-1977.