文献中正交试验的常见问题分析和解决方法

目的 提高正交试验的质量和实用性。方法 对文献报道的正交试验进行总结分析。结果 部分正交试验存在的问题和后果是：无空白列导致夸大随机效应；无理由忽略交互作用导致混杂现象；手工计算导致方差分析错误；最优工艺不考虑成本因素导致无谓增加生产成本。结论 正交试验应该留有空白列；否定交互作用的存在应该有实验或文献的支持；应该使用专业统计软件分析数据以确保结果的准确性；工艺考察应包括成本-效益评估。

Common Mistakes Made in Some Literatures When Performing Orthogonal Experiments and Solutions

OBJECTIVE To improve the quality and utility of orthogonal experiments. METHODS The orthogonal experiments in published literatures were reviewed. RESULTS The common mistakes made in some literatures and outcomes are: (1) orthogonal design or orthogonal table without blank column, resulting in inflating the random effect, (2) baselessly ignoring the interactions, resulting in confounding bias, (3) wrong results of Analysis of variance (ANOVA) because of manual calculating, or(4) the selection of optimal process without considering cost-effectiveness, incurring cost increase of manufacturing. CONCLUSIONS (1)Blank column of orthogonal table should be reserved for error estimated, (2) The interactions could be ignored only when their existence are denied in relevant experiments and literatures, (3) The data of orthogonal experiments should be analyzed using statistical software to make sure the accuracy of the results, and (4) The cost-effectiveness should be considered in the selection of optimal process.