关于正交试验设计的重复试验问题  

On duplicate test of orthogonal experiment design

作　　者：管宇[1] 黄必恒[1] 吴志松[1] 徐群芳[1] 

机构地区：[1]浙江林学院理学院,浙江临安311300

出　　处：《浙江林学院学报》2003年第4期413-418,共6页Journal of Zhejiang Forestry College

摘　　要：利用计算机进行大量随机模拟试验,揭示了在正交试验设计时因素水平差异显著与否,主要取决于试验效应的平方和与随机标准差,并推导出随机误差临界值近似计算公式:σ≈θ·r·m·∑α2k(a-1)·[Fα((a-1),fe)-1]。其正确性经随机模拟得以验证。建议做正交试验时尽可能设置2次重复试验。Utilizing computer to make a great scale of stochastic simulation experiments,this paper reveals that if the difference between levels of the primes exist noteworthily is mainly depended on the sum of its experiment effects and stochastic normal difference.An approximate calculation formula is derived:σ≈θ·r·m·∑α~ 2_k(a-1)·[F_α((a-1),f_e)-1].It is proved by stochastic simulation experiments.It is suggested that twice-experiments should be done at the orthogonal experiment.[Ch,3 fig. 3 tab.10 ref.]

关 键 词：正交试验设计 随机模拟 试验效应 随机误差临界值 

分 类 号：S11[农业科学—农业基础科学] S711