Logistic Regression Using SAS®: Theory and Application

9.6. Adjustment for Varying Time Spans

The Poisson distribution (or its generalization to the negative binomial) is well suited for describing counts of events that occur in some interval of time. In the preceding example, articles and citations were counted over a certain period of time. Other examples include the number of arrests in a five-year period, the number of colds that occur in one year, or the number of arguments between spouses in one month. If the length of the time interval is the same for every individual in the sample, the methods already described work just fine. But if events are counted over different lengths of time for different individuals, there is clearly a need for some kind of standardization. For ordinary regression ...