This Article  • Full text  • PDF  • Reply to article  • Send to a friend  • Save in My Folder  • Save to citation manager  • Permissions  Citing Articles  • Citing articles on ISI (4)  • Contact me when this article is cited  Related Content  • Similar articles in this journal

	Since this article does not have an abstract, we have provided the first 150 words of the full text and any section headings.

The landmark study of Steadman et al,¹ though carefully designed, has certain limitations. The investigators adopted the log-linear analysis as the principal method to analyze event rates for violent and aggressive behavior. This method ignores the fact that repeated observations are made in the same subjects, and assumes that measurements from the same subject are independent.

However, within-subject measurements over time are related. Omission of temporal relationships in repeated measures data can cause the probability of type I error to deviate from the nominal level of statistical significance (P<.05) and lead to false statistical inference.

Furthermore, as acknowledged by the authors, the substantial decline in violence over time might be confounded by a systematic attrition of patients during the follow-up period. Two procedures were employed to account for a systematic "attrition by patients who were violent": patients with incomplete data were excluded from the analyses and patients were . . . [Full Text of this Article]