Some Researches On Linear Regression Analysis

In this paper, we study the diagnosis of related linear regression method, the residual assumptions of statistical data analysis, variance stabilization transformation, the variance is to examine the state of the transform model is reasonable through, then we study the relevant statistic such as Cook statistic, WK Statistics, AP Statistics to explore the influence function, especially the study of the various Cook statistic values influence the selection of the model, we also discussed in the deleted data multiple regression effect size on the measurement problem. Finally, we study the outliers in the model to improve the model. We study two problems of regression diagnosis. First, we study the rationality of the model. If the model is unreasonable, we must improve the model to achieve our target.This paper is divided into three parts, the first part introduces the background; the second part examine whether the model reasonably, including chapters II and III. The third part research the problem of data fix, including chapters IV and V. The third part is base on the second part.