Estimation methods in linear regression models with proxy variables

In a regression analysis there are more than one explanatory variable that may affect dependent variable. All supported explanatory variables should be considered and included in the model for reliable results. However some explanatory variables supported by the theory are often unobservable. When an unobservable explanatory variable does not take part in regression model, it will lead to combine its effect of ε disturbance term and cause omitted variable bias. One solution for reducing this bias is using related proxy variable instead of unobservable variable. In this thesis linear regression model with proxy variable is considered and based on this model Alternative Ordinary Least Square Estimator, Inequality Constrained Least Squares estimator, Stein Rule Estimators of the parameter vector and The Iterative Estimator of the disturbance variance are examined.