| The approach is an empirical Bayesian methodology in a survival analysis context with an application to clinical trials. The intent of this proposed methodology is to give individuals in clinical trial equivalency research more flexibility in model selection. The methodology is an extension of Bartolucci and Singh's (1993) work to the Weibull and linear-exponential models that involves testing nonscale parameters by classical methods. For the nonscale parameter, the Thoman and Bain (1969) method is employed for the Weibull, and a nonconventional likelihood ratio test is derived in the linear-exponential case. Pertaining to the scale parameter, the approach defines a general class of discrepancy measures for equivalency, shows specific limiting cases of the general measures, and then applies the Bayesian neighborhood null hypothesis theory to derive posterior credibility regions on the measures for both distributions. |
