Parameter estimation methods in seemingly unrelated regressions

If the disturbance term of a regression equation is correlated with the disturbance terms of other regression equations in the same time point in an equation system with M multiple regression equations, this model is called seemingly unrelated regression (SUR) model. In this thesis study SUR model and its properties is defined and specifically M=2 case is analyzed. If explanatory variables of different equations in SUR model have multiple collinearity problem some alternative and efficient estimation methods are surveyed as an alternative to the OLS estimation method. These methods are also compared using mean squared error (MSE) criterion. For unknown variance-covariance matrix cases, feasible estimators and their properties are examined. Finally maximum entropy estimation, additivity property of disturbance terms in model and statement of the model with nonlinear functions are discussed.