Seminar :The Numerical Delta Method

Topic:The Numerical Delta Method

Speaker:Hong Han(Professor of Stanford University )

Time :Friday, 2:30-4:00 pm , June 23th, 2017


Language: Chinese  



This paper provides a numerical derivative based Delta method that complements the recent work by Fang and Santos (2014) and also generalizes a previous insight by Song (2014). We show that for an appropriately chosen sequence of step sizes, the numerical derivative based Delta method provides consistent inference for functions of parameters that are only directionally differentiable. Additionally, it provides uniformly valid inference for certain convex and Lipschitz functions which include all the examples mentioned in Fang and Santos (2014). We extend our results to the second order Delta method and illustrate its applicability to inference for moment inequality models. Our empirical application looks at inference for the maximum quantile treatment effect to determine which students benefit the most from a reduction in class size in the Tennessee STAR experiment.

Keywords: Delta Method, Bootstrap, Numerical Differentiation, Directional Differentiability


About the speaker :

Professor Hong graduated from Lingnan (University) College, Sun Yat-sen University in 1993 and got his PHD in Economics from Stanford University in 1998. As a professor of Stanford University, Professor Hong serves in many academy institutions: he is the fellow of the Econometric Society , the editor of the Journal of Econometrics , Journal of Business , Economic Statistics and the Econometrics Journal. He achieved a lot in the field of theoretical econometrics and applied econometrics ( especially the industry organization theory and the empirical method). Professor Hong’s theses are published on the top academic journals, such as Econometrica, Journal of American Statistical Association, Journal of Business and Economic Statistics, Review of Economic Studies, Journal of Econometrics, which have great influence in the academic community. Professor Hong also won the annual thesis prize of Journal of Econometrics for his essay: An MCMC Approach to Classical Estimation. Moreover, Professor Hong is the manager of several research projects supported by National Science Foundation and enjoys a high reputation in the academic community.

Research field : Econometrics, Industrial Organization , Applied Microeconomics


[Print] [Top] [Close]