A Robust Method for Testing Transformations to Achieve Approximate Normality
针对线性模型中实现近似正态性的幂变换，提出一种新方法，在蒙特卡洛实验中比似然法更稳健，比显著性检验法更有效。
Summary We propose a competitor to likelihood and significance methods for power transformations to achieve approximate normality in a linear model. The new method is shown in theory and a Monte Carlo experiment to produce more robust inferences than the likelihood method and more powerful (although possibly slightly less robust) inferences than the significance method.