In this chapter we consider several superpopulation models of victimization and derive an estimate of θ on the basis of each. The fit of each model is tested using a X2 goodness of fit statistic for each of the years 1973 though 1975. We find that a correlated Bernoulli model is an inappropriate model of crime, while a homogeneous Bernoulli model and a Markov model provide better fits to the data. None of these models fits the data as well as the model under which the modified ad hoc estimator is consistent.
Unable to display preview. Download preview PDF.