Geometry of the systems of the linear regression equations

Is represented the geometric interpretation of the method of least squares, which makes it possible to classify the possibility of the identification of the parameters of the system of linear regression equations. This gives the possibility in the cases of nonidentifiability or superidentifiability to advance the additional conditions, which guarantee the single-valued identification of the parameters of both the separate equation, and the entire system as a whole. An exponential increase in the number of versions of the partition of many factors into the endogenous and predetermined being investigated complicates the task of finding the regression interrelations. In this case the geometric interpretation of the system of regression equations makes it possible to goal-directed select the structural form of model, which ensures the significance of the parameters of all equations.