Hypothesis Testing of Multiple Inequalities: The Method of Smoothed Constraint Chaining
2010/07/27
研討會日期 : 2010-07-27
時間 : 14:30
主講人 : 陳樂昱助研究員
地點 : B110
演講者簡介 : Le-Yu Chen got his Ph.D. in Economics from University College London in 2009. He is currently serving as Assistant Research Fellow of Institute of Economics, Academia Sinica. His research interest is in econometrics, applied econometrics and microeconometrics.
演講摘要 : Econometric multiple inequality hypotheses arise in diverse ways. Examples include concavity restrictions on techological and behavioural functions, monotonicity and dominance relations, bounds on parameters partially identified via moment inequalities and orderings of predictive performance measures for competing models. The present paper contributes to research directed at making the testing of such hypotheses as routine in empirical work as the testing of equality restrictions. The key idea is to chain any given set of multiple inequalities into the form of an approximate single equality using smoothed zero-one indicator functions. We investigate the statistical properties of this form when parameter estimates are inserted. We are led to discover a consistent and asymptotically exact test procedure having the following advantages： (1) The test is pointwise similar on the boundary. (2) The test is unbiased against a core category of local alternatives where previous LF based tests fail. (3) The test's asymptotic null distribution, whenever non-degenerate, is standard normal and thus critical value of the test is computationally simple and can be a priori fixed.
