Escalation of Commitment and Nonorthogonal Analysis of Variance Revisited: A Comment on Schoorman, Bobko, and Rentsch

Abstract

Schoorman, Bobko, and Rentsch (1991) reanalyzed previously published data (Schoorman, 1988) and claimed that, in contrast to the original analysis, the data supported Schoorman's negative escalation hypothesis. We contend that this reanalysis was conceptually and technically flawed, and that Schoorman's original analysis yielded the correct (negative) conclusion. Schoorman et al. also discussed a variety of alternatives for conducting a factorial analysis of variance with unequal cell sizes, demonstrating that different procedures can lead to different statistical conclusions. However, their discussion is misleading and overlooks recent literature that has removed much of the mystery from the unequal‐n problem. We offer a brief review of the fundamental problems, and their solutions, in analysis of variance with unequal ns.