國立臺灣大學計量理論與應用研究中心 (CRETA)、中央研究院經濟研究所及臺灣經濟計量學會 (TES) 將共同舉辦 4 月份WETA @TES。4 月份 WETA 研討會相關資訊如下。

Date: April 22 (Friday), 14:00 pm – 15:00 pm

Venue: Raphael, GIS NTU Convention Center (集思台大會議 中心拉斐爾廳 (台北市羅斯福路四段 85 號 B1)

Topic: Recent Advancements in Nonlinear, Nonparametric,and Nonseparable Econometric Models with Measurement Errors

[Regirstration Fee]免費參加, 歡迎 4/18 (一)中午前踴躍報名參加!!

[About the Speaker]

Professor Yuya Sasaki is currently Assistant Professor of Economics at Johns Hopkins University. Professor Sasaki is an econometrician producing empirical methods for cross section and panel data. His research focuses on semi- and non-parametric identification of economic models with unobserved heterogeneity, measurement errors, and endogenous selection.

[Lecture Overview]

In this lecture, I present recent econometric techniques for nonlinear, nonparametric, and nonseparable models with measurement errors. The lecture starts with empirical evidence that motivates us to account for the possibility of measurement errors in econometric analysis. This is followed by a brief review of the linear models with measurement errors. This review illustrates among others the convenient fact that we do not need to use distributional features to estimate linear models.

As distributions play an important role for nonlinear and nonparametric models on the other hand, we next learn deconvolution techniques that are used to recover nonparametric distributions of the unobserved true variable from observed measurement(s). Two important cases are considered for deconvolution: (1) the case where the distribution of an error variable is assumed to be known; and (2) the case where the distribution is unknown, but two measurements are available.　Regularized estimation of the recovered density and its asymptotic convergence rates are presented for each case. With the recovered distributions for the unobserved true variable and/or the error variable, we discuss how to identify and estimate nonlinear and nonparametric models with errors in variables.

Another step forward is to consider nonseparable models. We study the spectral decomposition approach that allows us to nonparametrically identify non-separable models of measurement errors. We first consider the case where three measurements are available. This is followed by discussions of conditions under which two measurements suffice. This area is a twilight zone of econometric research in measurement errors.

I remark on relations between the measurement error modeling and various economic modeling techniques, such as mixture models and models with unobserved heterogeneity. The lecture concludes with examples of empirical applications of these econometric techniques. Cases are drawn from auctions, earnings dynamics, and education.

[Program]

April 22 (Friday ) Raphael, GIS NTU Convention Center (集思台大會議中心拉斐爾廳)

13:30-14:00: Registration

14:00-15:30: Lecture

15:30-15:45: Coffee Break

15:45-17:00 Lecture and Discussion

*Lecture in English