Testing for Normal Errors in Designs with Many Blocks
针对拟合参数较多的实验设计，提供了检验同方差正态误差假设的拟合优度检验，给出了Cramér-von Mises、Watson和Anderson-Darling统计量的渐近临界点，并通过蒙特卡洛研究辅助使用。
Goodness-of-fit tests are provided for the assumption of homoscedastic normal errors in experimental designs where the number of fitted parameters is large. Asymptotic critical points are given for the Cramér-von Mises statistic, Watson's statistic and the Anderson-Darling statistic. An expansion of the covariance function of the empirical process of standardized residuals is given. The corresponding weak convergence result is established for one-way layouts when the number of parameters grows linearly with the sample size. A Monte Carlo study is given to aid in the use of the tables.