Application of the disposition model to breast cancer data

In this paper, we have presented the second level nesting of Bonney's disposition model (Bonney, 1998) and examined the implications of higher level nesting of the disposition model in relation to the dimension of the parameter space. We have also compared the performance of the disposition model with Cox's regression model (Cox, 1972). It has been observed that the disposition model has a very large number of unknown parameters, and is therefore limited by the method of estimation used. In the case of the maximum likelihood method, reasonable estimates are obtained if the number of parameters in the model is at most nine. This corresponds to about four to seven covariates. Since each covariate in Cox's model provides a parameter, it is possible to include more covariates in the regression analysis. On the other hand, as opposed to Cox's model, the disposition model is fitted with parameters to capture aggregation in families, if there should be any. The choice of a particular model should therefore depend on the available data set and the purpose of the statistical analysis.