More on Score Tests for Transformation in Regression
将Lawrance(1987)关于回归中变换的检验结果扩展到响应变量和均值均可变换的情形，并指出两种方法得到的统计量近乎完全相关，且观测信息与期望信息作为得分统计量的标准化效果差异显著。
Lawrance's (1987) results concerning tests for transformation in regression are extended to models where response and mean may both be transformed. It is also suggested (i) that the methods of Lawrance and of Atkinson (1982) lead to statistics with near-perfect correlation, and (ii) that observed and expected informations may have considerably different effects as standardizations of the score statistic.