一种新的统计回归模型及其建模方案

A New Statistical Regression Model and Its Building Method

根据经典门限自回归模型的基本思想，引人半截多项式变换，导出了一种新的统计回归模型，并提出了相应的一整套建模方案。这种模型和建模方案的特点是：（１）解决了统计回归分析中逐段线性化模型的检验问题；（２）在确定统计回归模型中各待估参数和变量──包括门限变量、门限值、阶数、时滞和回归系数时显得十分方便、快捷。试验结果表明，根据此方案建立的统计回归模型具有较高的拟合和预报精度，同时具有良好的稳定性。

On the basis of principle of the classical threshold auto--regression model, we advanced anew statistical regression model and a relative building method via introducing the semi--polynomialtransformation. Its characteristics includes' First, solving the test problem of stagewise linear model inthe statistical regression analysis. Second, determing the unestimated parameters and variablesconveniently and promptly. The results show that the new statistical regression model is more efficientand stable than the classical one.