A frequent problem in estimating logistic regression models is a failure of the likelihood maximization algorithm to converge. In most cases, this failure is a consequence of data patterns known as complete or quasi-complete separation. For these patterns, the maximum likelihood estimates simply do not exist. In this paper, I examine how and why complete or quasi-complete separation occur, and the effects they produce in output from SAS ® procedures.I then describe and evaluate several possible solutions.

The first three experiments are part of Jeffrey A. Hadley's doctoral dissertation at the University of Colorado (Boulder), performed under the guidance of Alice F. Healy. We thank Philip Langer for his help as cochair of the doctoral committee, Antoinette Gesi for her help conducting and analyzing Experiment 4, and Albrecht Inhoff and Marilyn Smith for their helpful comments concerning earlier versions of this article.

No abstract

No abstract

Drinking increases the RR of dying while boating, which becomes apparent at low levels of BAC and increases as BAC increases. Prevention efforts targeted only at those operating a boat are ignoring many boaters at high risk. Countermeasures that reduce drinking by all boat occupants are therefore more likely to effectively reduce boating fatalities.