Aggregation Bias and Ecological Fallacy
Researchers need to design their studies at the appropriate unit of analysis. This is especially important to understand so that one can avoid making improper inferences about units of analysis that were not actually analyzed. Confusing the units of analysis can lead to an 'ecological fallacy', where one attributes an association between variables to operate at a lower level of disaggregation than actually studied. Often group level statistics reflect an aggregation bias compared to individual or lower level statistics. For example, one commits an ecological fallacy when one concludes that individuals from families in poverty do less well in school by studying the correlation between the average poverty level in schools and school test averages. These issues are now studied using 'multi-level', or 'mixed' statistical models that take into account different, nested units of analysis.
 |
 |
|
|
