Abstract
We compare the two-part decision making hypothesis that distinguishes between users and non-users with the one-part decision making hypothesis of medical care demand. Using count data finite mixture techniques, this paper provides more flexible frameworks for medical care demand that have features of both hypotheses.
We apply the deterministic annealing EM algorithm in order to estimate complicated finite mixture models. In model comparison using information criterion and goodness of fit tests, we find some evidence that the frequently used one-part finite mixture model may not be adequate to describe this type of medical care demand and that the one-part and two-part finite mixture model may be more desirable.
