计量分析 期中考试.pdf

Econ 507. Econometric Analysis Walter Sosa-Escudero Midterm Exam Rules: you may use one letter-sized sheet with your own notes and a calculator. Unless stated otherwise, the notation corresponds to the one used in class. Each question is worth 25 points. Good luck! 1. Solve only one of the following two problems (do not solve both, if you do so, only the ?rst one you report will be graded.) (a) Prove that the two-stage least squares estimator is asymptotically normal and derive its asymptotic variance. Use the asymptotic framework utilized in class, for the p K case (more instruments than regressors) but do not assume conditional homoskedasticity. You may assume consistency. Show all the steps clearly and the assumptions you use in each step. (b) Consider the linear model under homoskedasticity and/or serial correlation, so that V (u|X) = σ2?. i. Prove that σ2(X0X)?1(X0?X)(X0X)?1 ? σ2(X0??1X)?1 is positive semide?nite ii. Intuitively, what does this result tell you? 2. True of False: ?rst indicate whether each of the following statements are true or false, and then justify your answer briey, in no more than 5 lines per item. No points will be given to questions without justi?cation or with incorrect justi?cation. Unless indicated otherwise, the notation corresponds to the one used in class. (a) The feasible GLS estimator is the best linear unbiased estimator of the parameters of the linear model under heteroskedasticity of serial correlation. (b) Under conditional homoskedasticity the two-stages least squares estimator is asymptoti- cally more e–cient than the optimal GMM estimator. (c) The assumption of normal errors is necessary to show that the Wald statistic for H0 :