The Analysis of Transformed Data
针对Bickel和Doksum（1981）提出的数据变换后需大幅修改标准分析方法的观点，本文基于Box和Cox（1964）的工作，从参数解释和贝叶斯分析角度论证相反结论，并用数值例子说明。
Abstract Recently it was suggested (Bickel and Doksum 1981) that when data are used to select a transformation, the post-transformation analysis of those data may need to be modified considerably from standard form so as to allow for the selection. We argue that common sense and the work of Box and Cox (1964) point to a contrary conclusion. Our argument is based on considerations of parameter interpretation and subsequent Bayesian analysis, within the context of fitting normal-error linear models. Numerical examples are used to illustrate the main points.