最优Box-Cox变换模型的反向动态规划求解

Reversely solution to optimal box-cox transformation modeling

通过构造Box Cox变换模型加权最优目标函数,综合考虑回归系数t统计量对应p阀值、Theil系数U2及其误差分解、修正回归拟合优度、Durbin Watson检验量、回归残差。利用Excel矩阵向量数组函数三维动态建模,通过Excel基于稳健GRG2算法的规划求解,"反向"动态求解整体最优Box Cox变换模型估计参数。实例给出了建模过程,显示反求建模可较有效地克服变换模型的异方差、多重共线性和序列相关性。

A weighted target optimal funtction for BoxCox transformation is constructed including the regression t and F tests, Muticollinearity and Heteroscedasticity. By using Microsoft Excel array functions, the dynamic spreadsheet model calls initial robust GRG2 Solver to reversely seek global optimizing parameters of BoxCox transformation. The example illustrates the modeling course and the satisfying application for multicollinearity and autocorrelation in regression.