Lecture4

# Under the null hypothesis and using the variance of

• Notes
• 17

This preview shows page 10 - 12 out of 17 pages.

) under the null hypothesis, and using the variance of this difference to determine whether or not a given-sized difference is significant. If the difference is significant at some chosen level of test size, we reject the null hypothesis. If there is a significant difference between the two sets of estimates, one can regard this as evidence supporting the hypothesis that there is correlation between regressors and the disturbance. In such a case, OLS is inconsistent, and GIVE should be used. 6.2 THE WU-HAUSMAN TEST The Hausman test can not be implemented in all econometrics packages. A similar procedure which can be done in almost any package makes use of the Wu-Hausman test. This is sometimes described as an “exogeneity” test, but is more properly thought of as a test of regressor “orthogonality”; that is, a test of the assumption that there is zero asymptotic correlation between the regressors and the disturbance term. The null and alternative hypotheses are again H 0 : plim 1 T X u = 0 H a : plim 1 T X u 0 The test is implemented in a series of steps. For simplicity, in the presentation which follows we suppose that only one of the regressors might be asymptotically correlated with the disturbance. Suppose that this is the variable X 4t in the regression model: Y X X X u t t t t t = + + + + β β β β 1 2 2 3 3 4 4 We also assume that the researcher has available two instruments, W 1 and W 2 , for the potentially correlated regressor X 4 . STEP 1 : Estimate by OLS the “ instrumenting regression ”, and save the residuals. This is the regression of the potentially correlated regressor on all instruments. Note carefully that by “instruments”, we mean here all explanatory variables in the original regression model that are not being treated as potentially correlated, together with all additional variables being used as instruments for the potentially correlated variable. So in our example, the “instrumenting regression” to be estimated by OLS is X X X W W t t t t t t 4 1 2 2 3 3 4 1 5 2 = + + + + + α α α α α ε Denote the saved residuals as ε t , for t=1,...,T. STEP 2 : Estimate by OLS the original regression model: Y X X X u t t t t t = + + + + β β β β 1 2 2 3 3 4 4 STEP 3 : Use a variable addition test to test for the significance of the residuals when added as a regressor to the original regression model. That is we wish to test H 0 : θ = 0 against H A : θ 0 in the OLS regression

Subscribe to view the full document.

11 Y X X X u t t t t t t = + + + + + β β β β θε 1 2 2 3 3 4 4 If we reject H 0 , we regard this as evidence in favour of the alternative hypothesis that X 4t is correlated (asymptotically) with the disturbance term, and so should use GIVE rather than OLS to obtain our parameter estimates. If we cannot reject the null, the OLS estimator is preferred.
You've reached the end of this preview.

{[ snackBarMessage ]}

### What students are saying

• As a current student on this bumpy collegiate pathway, I stumbled upon Course Hero, where I can find study resources for nearly all my courses, get online help from tutors 24/7, and even share my old projects, papers, and lecture notes with other students.

Kiran Temple University Fox School of Business ‘17, Course Hero Intern

• I cannot even describe how much Course Hero helped me this summer. It’s truly become something I can always rely on and help me. In the end, I was not only able to survive summer classes, but I was able to thrive thanks to Course Hero.

Dana University of Pennsylvania ‘17, Course Hero Intern

• The ability to access any university’s resources through Course Hero proved invaluable in my case. I was behind on Tulane coursework and actually used UCLA’s materials to help me move forward and get everything together on time.

Jill Tulane University ‘16, Course Hero Intern