Is Victimization Chronic? A Bayesian Analysis of Multinomial Missing Data

Study data came from interviews conducted 6 months apart in a rotation design. Bayesian methods were used to analyze the data and to explicitly account for the fact that not all the intended interviews could be conducted. The Bayesian approach was computationally feasible for some, but not all, of the calculations. Study results indicate that victimization is chronic, in that there is a factor of about 3.7 greater odds of a household once victimized being victimized again. A deeper analysis of victimization would look for household characteristics that explain victimization, in the sense that, given those characteristics, victimization is not chronic. Socioeconomic status, or some surrogate of it, might be a good first variable to use. Appendixes discuss the computational methods, and a list of 10 references are supplied.