回归模型中的变量变换及其应用

Transformations to Variables in Regression Models and its Application

按回归模型的数学形式,可以把回归模型划分为线性模型和非线性模型。而对于非线性模型,函数形式的确定是一大难题。以往的方法是假定某一初等函数去拟合样本数据,然后用线性化代换或Newton迭代法去估计模型参数。这种全凭经验和视力来判断属于何种函数类型的方法,本身就存在一定的误差;

This paper describes a nonliear transformation(BOX—COX) to variables in regression models. One approach is introduced also to estimation for transformation models. The ideal results are obtained through the calculation of examples and show that it has broad prospects to apply the method in practice.