同时处理多种回归异常的一般方法
A General Approach to Handling Multiple Regression Exceptions at the Same Time

回归分析是以概率论与数理统计为基础，主要对随机现象统计资料进行分析和推断。而在实际的应用中发现同时满足基本假设的数据是非常少的，通常会出现多重共线性，异方差性和自相关性的问题。然而大多数的教材都只是给出了出现单一的问题时的解决办法，因此本文为了解决当一个回归分析中出现上述三个问题中的两个或者三个的时候应当以何种顺序解决的问题，采用了理论分析和实例验证的方法。从实例和分析结果来看当同时出现违背基本假设的多种情况下，如果数据为截面数据时的处理顺序是多重共线性–异方差性–自相关性，当数据为时间系列数据时的处理顺序为多重共线性–自相关性–异方差性。
Regression analysis is based on probability theory and mathematical statistics, and mainly analyzes and infers the statistical data of random phenomena. In practical applications, very few data are found to meet the basic assumptions at the same time, and there are often problems of multicollinearity, heteroscedasticity and autocorrelation. However, most textbooks only give solutions to a single problem, so this paper uses theoretical analysis and example verification methods to solve the problems in what order two or three of the above three problems should be solved in a regression analysis. From the example and analysis results, when there are multiple cases that violate the basic assumptions at the same time, the processing order is multicollinear-heteroscedasticity-autocorrelation when the data are cross-sectional data, and the processing order is multicollinear-autocorrelation-heteroscedasticity when the data are time series data.