Grouped Likelihood for the Shifted Power Transformation
针对移位幂变换模型，普通似然函数表现不佳（常无局部最大值），本文提出组似然方法可避免这些问题，并讨论了实际应用；贝叶斯推断也面临类似困难。
SUMMARY Estimation of parameters is considered in the shifted power transformation family of models. The ordinary likelihood function is shown to behave poorly, often having no local maximum. It is argued that a grouped likelihood approach avoids these difficulties, and the practical implementation of this approach is considered. Bayesian inference is also considered, though more briefly, and it is argued that there are again difficulties with an ungrouped likelihood approach. The results provide further evidence of the utility of grouped likelihood approaches to inference in non-regular problems.