Parameters Estimations In PANEL Models

Panel data model is an important linear model in economics,finance, biology,medicines and other fields.In recent twenty years,statistical in-ferrence about this model attracts many statisticians.In this paper,we first generalize the latest development of parameter estimation in this field,then focus on parameter estimation in the Panel model with individual effect and time effect.Many articles researched the parameter estimation of the regression coefficents in the case that both individual effect and time effect are random,but in some conditions,it is more reasonable if we suppose either of them is fixed.This paper is based on this hypothesis to research the estimations of the coefficents .The variance-covariance matrix still include parameter of variance in this condition,so our purpose is to look for feasible estimations. Through proper linear transformation of the origional model, we obtain the educed models which are all sigular models.We peform the uniform theory of least squares getting the best linear unbiased estimates of the coefficents and the ordinary least squares estimates. All these estimates are feasible unbiased estimates.We use the quotient of the determinant of the variance-covariance matrix of the feasible unbiased estimates,that is ,a kind of relative efficency to compare these esti-mates.The results have instructive significance for practice.