الفهرس | Only 14 pages are availabe for public view |
Abstract Model Estimation is the second step in any statistical analysis. It is very important since all the proceeding steps depend on its efficiency. The most popular estimation technique in linear regression analysis is the least squares method which under certain assumptions results in the family of best linear unbiased estimators. However, if we considered relaxing one of the assumptions to use a model that has broad application in some economic phenomena then it{u2019}s possible to obtain a family of biased estimators which is nonlinear function of observations on the dependent variable and has a smaller mean squared error. The model considered in the present thesis is a set of individual linear multiple regression equations which aren{u2019}t related in the variables but the contemporaneous disturbances associated with at least some equations are correlated with each other. That is, the equations may be linked statistically, even though not structurally, through the jointness of the error terms{u2019} distribution and through the non-diagonality of the associated variance covariance matrix. The model can be estimated equation-by-equation using standard ordinary least squares method (OLS) or jointly using Zellner (1962) method (SURE). The OLS method leads to consistent estimators, however generally not the most efficient estimators compared to as the SURE method. Zellner (1962) used the expression 2seemingly unrelated regression equations (SURE)3 as a way to express the special nature of this model in his study. In 1963, Zellner explored the finite sample properties of these estimators in two-equation regression models |