The Desirability of Covariance Adjustments
通过模拟研究比较两种调整准则，建议除非协变量与响应变量相关性极低或自由度严重不足，否则应调整协变量以估计处理效应差异。
The decision on whether or not to adjust the estimate of the difference between the effects of two treatments according to the value of a covariate is investigated by a simulation study. Two criteria for deciding whether or not to adjust are compared. Adjustment is recommended unless it is known that the covariate is very poorly correlated with the response variable or there is a serious lack of degrees of freedom for estimating the regression coefficient of the response variable on the covariate value. The reasons for this are discussed.