The study used data from the 1996-98 National Crime Victimization Survey (NCVS) to estimate repeat victimization rates for all violent crimes and separately for domestic violence. It examined repeat victimizations among individuals rather than households. It introduced two new models for incorporating information from partial records into estimation of repeat victimization rates. Further, it described the construction of a longitudinal data set from the NCVS and presents the missing data patterns in the longitudinal data. The study concluded that estimates of repeat victimization from the NCVS depended heavily on the form of the model used to impute values for missing data. One aspect of the models developed by this study was that the imputation depended in part on victimization status in the other year. Inclusion of status in the other time period served as a proxy for unobserved variables that might better predict repeat victimization. Tables, references