论文部分内容阅读
因子效应原则(效应稀疏原则、效应排序原则和效应遗传原则)经常用于评判因子设计理论与数据分析策略的合理性.针对非正态响应的部分因子试验,当筛选试验含有复杂的别名效应时.提出了一种结合广义线性模型(generalized linear models,GLM)与因子效应原则的多阶段贝叶斯变量选择方法.首先,在广义线性模型的线性预测器中对每个变量设置一个二元变量指示器;其次,将因子效应原则以变量指示器的先验信息分成三个不同的阶段分别加以考虑;然后,利用变量指示器的后验概率识别显著性的因子效应.最后,仿真试验结果表明:所提出的方法不仅能简化广义线性模型先验参数的选择,而且能够有效地识别出非正态响应部分因子试验的显著性因子.
The principle of factorial effects (sparsity, effect and genetics) is often used to evaluate the rationality of factorial design and data analysis strategies. A multi - phase Bayesian variable selection method combining generalized linear models (GLM) and the principle of factor effects is proposed.Firstly, a binary variable is set for each variable in the linear predictor of generalized linear model Secondly, the factor effect principle is divided into three different stages according to the priori information of the variable indicator, and then the significant factor effect is identified by the posterior probability of the variable indicator.Finally, the simulation results show that : The proposed method can not only simplify the selection of a priori parameters of generalized linear models, but also effectively identify the significant factors of some non-normal response factor tests.