人民大学高级计量经济学讲义11.ppt

Econometric Theory Lecturer: Dr.Jingtao Yi Room 715 Business School, RUC Lecture 11: k-Variable Linear Regression VIIII 1. Specification Error; 2. Tests of Parameter Constancy; 3. Tests of Structural Change. Specification Specification Error Problems with Functional Form Specification Error Problems with b Specification Error Problems with u Specification Error Problems with X Test of Specification Error The Ramsey RESET Test Tests of Parameter Constancy In an estimated equation, parameters are assumed to be constant over all observations in the sample: “Structural stability”. 1.Tests of Structural Change The estimated equation has relevance for data outside the sample data used in the estimation, which is embodied in the notion of parameter constancy: “Predictive accuracy” or “Out-of-sample prediction”. 2.Tests of Predictive Accuracy Tests of Structural Change Tests of Structural Change Tests of Structural Change The test procedure: 1. Using the n1 observations, regress y1 on X1; using the n2 observations, regress y2 on X2; obtain the unrestricted RSS, e1’e1+e2’e2; 2. Fit the same regression to all (n1+n2) observations and obtain the restricted RSS, e*’e*; 3. Compute the F-stat and reject the null if F-stat exceeds the critical value. Tests of Predictive Accuracy The Chow Forecast Test Tests of Predictive Accuracy Tests of Predictive Accuracy The Chow test procedure: 1. Using the n1 observations, regress y1 on X1 and obtain the unrestricted RSS, e1’e1; 2. Fit the same regression to all (n1+n2) observations and obtain the restricted RSS, e*’e*; 3. Compute the F-stat and reject the null if F exceeds the critical value. * *