Czobor and Volavka's comments contain 3 main points. First, they are concerned that we used a log-linear model to analyze differences in aggregate rates of violence and that we ignored within-subject variability. Second, they hypothesize that the observed downward trend in postdischarge violence rates is due to the "biased attrition" of a certain group of patients "with persistent, recidivistic violence." Finally, Czobor and Volavka are concerned that our treatment of "time at risk" as a covariate in attempting to explain the observed decline in violence violated an assumption of independence.
Steadman HJ, Robbins PC, Silver E, Mulvey EP, Roth LH, Monahan J, Appelbaum PS, Grisso T. Violence in the Mentally Ill: Questions Remain—Reply. Arch Gen Psychiatry. 1999;56(2):193-194. doi: