Welcome Tp CRETA! Contact Us at : +886.2.3366.1072

Welcome To CRETA

國立臺灣大學計量理論與應用研究中心 - CRETA



2010 年 11 月份第二場 WETA


2010 年 11 月份第二場 WETA


WETA@TES November 2010 II - 26 November 2010, Kuan Te Lecture Hall, Bldg. 1, College of Management

由於 10 月 30 日為臺灣經濟計量學會年會,10 月份 WETA 暫停一次。11 月份則將舉行兩場 WETA,分別邀請到 University of Essex 的 Professor Giorgio Valente (11 月 19 日) 及中央研究院經濟研究所的陳宜廷博士( 11 月 26 日)至 CRETA 進行演講。

【2010 年十一月份第二場 WETA 研討會】

日期:2010 年 11 月 26 日
地點:臺灣大學管理學院一號館 2F 冠德講堂
主講人:陳宜廷博士 (中央研究院經濟研究所)
時間:14:00~15:15 session 1
15:15~15:45 茶 敘
15:45~17:00 session 2
講題:Maximum Entropy Principle: Review and Applications

The maximum likelihood (ML) method is known as the best statistical inference method in the case where the true data generating process (DGP) is known. Many parametric specification, estimation, and testing methods explicitly or implicitly claim their optimality following the ML principle. However, the fact is that the true DGP is unknown. A more realistic situation is that we could only learn partial information about the real world either from economic theories or statistical observations. Put differently, although the ML principle is a golden rule in theory, it is infeasible in practice. This fact has considerably motivated the use and development of the method of moments (MM) and its extensions and variants, like the generalized MM (GMM) and the quasi-ML(QML) methods, in econometrics. A common feature of these robust methods is that they do not rely on, and hence do not pursue, a complete (conditional) distribution specification for parameter estimation. However, we do need a complete distribution specification in many economic and financial problems. In this scenario, the maximum entropy (MaxEnt) principle is useful because it allows us to recover a distribution specification from a set of data-consistent, or theory-consistent, moment conditions in a "least-biased" way.
In the first part of this lecture, we will review some key concepts and appealing properties of the MaxEnt principle, discuss the associated implementation issues, and provide personal discussions about this approach.
In the second part, we will discuss some existing econometric applications and introduce personal studies of this principle.


陳宜廷教授為臺灣大學經濟學博士,目前任職於中央研究院經濟研究所,研究領域為 Econometrics, Time Series Analysis, Empirical Finance,詳細期刊論文著作請參閱 陳教授網頁 http://www.econ.sinica.edu.tw/research01.php?researchID=24&foreLang=tw&secureChk=7e3988810e274b35a3f4ed1a1bca90c7

議  程:

14:00~15:15 session 1
15:15~15:45 茶 敘
15:45~17:00 session 2