论文部分内容阅读
ｌｏｇｉｓｔｉｃ回归模型的目的是描述因变元与自变元之间的关系，回归系数有明确的实际意义。一旦回归系数的估计受非典型数的干扰，就难于获得对实际问题的解释。文中讨论了解释变量多元共线对ｌｏｇｉｓｔｉｃ回归系数影响，在此基础上引用线性回归系数主成分估计的思想，改进了ｌｏｇｉｓｔｉｃ回归系数的加权最小二乘估计，使之能克服多元共线引起的一般ｌｏｇｉｓｔｉｃ回归系数加权最小二乘估计方差扩大现象。文中通过模拟数据说明了主成分改进的加权最小二乘估计较一般加权最小二乘估计优越
The purpose of the logistic regression model is to describe the relationship between the variable and the self-variant, and the regression coefficient has a definite practical significance. Once the estimate of the regression coefficient is disturbed by the atypical numbers, it is difficult to obtain an explanation of the actual problem. This paper discusses the influence of multivariate collinearity on the logistic regression coefficients of explanatory variables. Based on this, it cites the idea of principal component estimation of linear regression coefficients, and improves the weighted least squares estimation of logistic regression coefficients so that it can overcome the general problems caused by multicollinearity. The expansion of the variance of the weighted least squares estimate of the logistic regression coefficients. The simulation data shows that the weighted least squares estimation of principal component improvement is superior to the general weighted least squares estimation.