Courts are paying increasing attention to the scientific validity of regression analyses. Regression analyses have long been used in securities fraud cases in the form of event studies that purport to determine the extent to which misrepresentations caused investor losses. Regression analyses also are used in employment discrimination cases, primarily due to the lack of direct evidence of discrimination and the resulting need to rely on circumstantial evidence. More recently, regression analyses have been used in consumer class actions to test for differences (or similarities) across the putative class.
Issues of sample size and validity, omitted variables, and statistical significance often are central to the debate over the validity of regression results. Proper resolution of these issues can be the difference between a reliable and resilient analysis and one subject to attack.