中国烟草学报

以文献的试验数据为例,较为系统地探讨了多指标正交试验数据的优化分析方法。结果表明:(1)多指标正交试验的直观分析最好要考虑误差(空白列)的影响和进行方差分析,以准确寻找重要试验因素;(2)对于试验效应不显著的因素,不宜与评价指标做相关性分析;(3)不同评价指标下各因素的最优水平存在一定的差异,应综合方差分析、因素贡献率和同一因素不同水平效应差异显著性检验的结果,确定试验最优水平组合。

Analysis and optimization of data from orthogonal design was carried out.Results showed that:1) range analysis of multi-index orthogonal experiment should be included in the analysis of blank list,analysis of variance should be performed in order to determine important experimental factors;2) correlation coefficient of a factor and experimental indicator needn't be calculated when the effect of the factor is non-significant;3) according to different evaluated index,the optimal level of a factor differs and the optimal levels of different factors may be decided by analysis of variance and contribution ratio of the factor's and multiple comparisons of one factor at different levels.