Portal de Periódicos da CAPES

Two Sources of Error in Ecological Correlations

1973; SAGE Publishing; Volume: 38; Issue: 6 Linguagem: Inglês

10.2307/2094137

1939-8271

John L. Hammond,

Species Distribution and Climate Change

The discrepancy between individual and ecological correlations has often been noted, but its sources have not been understood. The discrepancy can arise from two quite distinct sources, both of which can be explained sociologically. The two sources have opposite implications for the possibility of inference to individual relationships from aggregate data. When individuals are grouped into neighborhoods on the basis of their homogeneity on an independent variable, the ecological correlation will necessarily be larger than the individual correlation; but the regression equation of the aggregate variables provides an unbiased estimate of the individual regression. Aggregation bias arises when the independent variable has a contextual effect, or when individuals are grouped into neighborhoods on the basis of their similarity on the dependent variable. If aggregation bias is present, no inference about the individual relationship can be drawn from aggregate data. An investigator's knowledge of the social processes operating in the situation he is examining will often enable him to estimate whether his data incorporate aggregation bias; if they do not, he can draw inferences about the individual relationship from aggregate data.