|
Note on a Large-Sample Result in Specification Analysis
T. Kloek
Abstract
If two linear models have different sets of explanatory variables and the same variable to be explained, the residual variance of the correct model (S^2"n) has a smaller mean value than that of the incorrect one (t^2"n). This note shows under fairly general conditions that S^2"n < t^2"n will hold with probability arbitrarily close to 1 provided that the sample size n is large enough.
|
