Extreme Groups Designs

An extreme groups design is one in which a researcher samples cases from the extremes of the distribution of an independent variable X and then examines the relationship between X and another variable Y. Extreme groups designs can lead to significant cost savings in data collection while preserving or enhancing statistical power for detecting relationships. However, methodologists have noted several limitations associated with this method that often go unrecognized, including artificially inflated effect size and compromised reliability. In addition, use of such designs often requires the researcher to make additional assumptions about the variables and how they are related. Use of extreme groups designs should be employed only after careful consideration of their advantages and disadvantages. Keywords: abnormal; acculturation; addiction; effect size; extreme groups; interaction effects; statistical power; subgrouping