Empirical studies of violence and mental illness have used many different methods. Current state-of-the-art methods gather information from both subject and collateral interviews as well as official records. Typically these sources are treated as additive. Any report of a violent incident from any source is treated as true and all reported incidents are added to generate estimates of frequency. This paper presents a new statistical technique that uses the level of agreement between the sources of data to adjust those estimates. The evidence suggests that, although the additive technique for using multiple sources correctly estimates how many people are involved, it substantially underestimates the number of incidents. The new technique substantially reduces both false negatives and false positives.