| Regression analysis is a very important part in statistics. Its core content arecoefficient evaluation and significance test. According to these questions, this articlestudies test method, the matrix expression of coefficient evaluation and significancetest, the diagnosing and cure of morbid matrix. In already existing literatures, somebasic principles and tackling methods are given to these questions, but there are stillmany places which can be amended. On these foundations, this article posed severalimportant complementarities and improvement.Article separately studies the amending of significance test in simple linearregression model by using repeating observing data to testing the quality of the model;happening, hazards and evading in different aspects to tackling co-linearity in linearregression, and finally poses a ameliorationFirst, the article studies the relation between sum of square value of regressionand sum of square value of errors by testing coefficient. And then article gives anamending method of simple regression analysis. Article gives a data example, makingresults by the software SAS, and then proves that this method is feasible.After that, as to the multiple regression model, which adds a independent variantupon simple regression model, by matrix we can know the frankness of the test andevaluation which we care. On this base we portrayed the test and cure method ofmorbid matrix, and studies the simple resolution and gives a example, representing thefeasibility of this amending.At last, article expounds the forming and hazard of co-linearity in multipleregressions, giving.an example of how to ensure co-linearity. On the base of tacklingthe co-linearity, article gives a easier amending method. At last there is a example, bythe analysis of SAS we assert the feasibility of amending method of the co-linearity. |
