近年来 ,大量研究致力于建立新的回归技术解决经典回归中假设过于严格的问题 ,包括预测变量与反应变量的线性关系、反应变量的正态性及方差齐性等。当线性关系不成立时选用线性模型明显不适宜 ,此时一种方案是在模型中加入交互作用项或采用变量变换 ,但这种方式容易导致解释时? In recent years, a large amount of research has been devoted to the establishment of new regression techniques to solve the problem of overly rigorous assumptions in classical regressions, including the linear relationship between predictors and response variables, the normality of response variables, and the homogeneity of variance. When the linear relationship is not established, it is obviously not appropriate to use the linear model. At this time, a scheme is to add an interaction term or a variable transformation to the model. However, this method can easily lead to interpretation.