Using the Statistical Concept of “Severity” to Assess the Compatibility of Seemingly Contradictory Statistical Evidence (With a Particular Application to Damage Estimation)

Abstract

When parties present divergent econometric evidence, the court might either combine such evidence in an ad hoc way or view such evidence as contradictory and thus ignore it completely, without conducting closer analysis of the possible sources of the contradiction. We believe that the reasons for this development are (i) that the statistical evidence is often interpretated in a simplistic manner and (ii) that the fact is ignored that any statistical test tests within the boundary of a prespecified model which might be wrong. Contradictory evidence might therefore either occur by chance or because the underlying assumptions contradict each other. Based on the concept of severity, we propose a method to avoid common fallacies in the interpretation of empirical evidence. We further set out a simple method for distinguishing between actual and merely apparent contradiction based on the statistical concept of the “severity” of the furnished evidence. Our chosen application is that of damage estimation in follow-on cases.