This thesis discussed the estimation of regression coefficient in the linear model. When we want to build a linear model for satisfying some practical need, there may be some covariates which are linearly related and enter the final model so that linear square estimate has large mean square error and loses its value in practical application. In the paper, How to improve the linear square estimate is discussed.Chapter 2 analyses the cause which creates collinear. From the practical cause to theory analysis, the mechanism of collinear is described and explored. The consequence of collinear for linear square estimation is given by the mean square error.Chapter 3 discuses ridge estimation. There must exist a parameter K so that the mean square error of ridge estimation is little than linear square estimation's.In Chapter 4, the mixture estimation of OME and ORE is put up. The novel optimal properties are explored. The example proved that the mixture estimation is better than the previous estimation and has a satisfied consequence for our data.
|