A CAUTIONARY NOTE ON THE ANALYSIS OF EXTREME DATA WITH COX REGRESSION

This article analyzes a small censored data set to demonstrate the pot ential dangers of using statistical computing packages without underst anding the details of statistical methods. The data, consisting of cen sored response times with heavy ties in one time point, were analyzed with a Cox regression model utilizing SAS PHREG and BMDP2L procedures. The p values, reported from both SAS PHREG and BMDP2L procedures, for testing the equality of two treatments vary considerably. This articl e illustrates that (1) the Breslow likelihood used in both BMDP2L and SAS PHREG procedures is too conservative and can have a critical effec t on an extreme data set, (2) Wald's test in the SAS PHREG procedure m ay yield absurd results from most likelihood models, and (3) BMDP2L ne eds to include more than just the Breslow likelihood in future develop ment.