演講主題:Review on Multiple Hypotheses Testing without Data Snooping Bias
講題摘要:
In a multiple hypotheses testing problem, it is often of interest to identify as many false null hypotheses as possible while accounting for the data-snooping effect. For example, among a given set of models such as portfolios, mutual funds, hedge funds or trading rules, one would like to examine if there are some models having superior performance relative to a benchmark. Then, data snooping may arise because, when many models are evaluated individually, some are bound to be superior by chance alone even though they are not.
In this talk, I will review several multiple hypotheses tests in the literature that are not subject to data snooping bias. We will also see how one can improve the power of a test by adopting different definition notions of error rates such as familywise error rate, k- familywise error rate and false discovery rate.
On the other hand, when one tests for the multiple inequality constraints and if the least favorable configuration (LFC) is adopted, the resulting test is conservative and less powerful. We introduce the recentering method proposed by Hansen (2005) that can improve the power of a test by avoiding using LFC.
講者介紹:
許育進教授為美國 University of Texas 經濟學博士,目前任職於中央研究院經濟研究所。研究專長為經濟計量理論,詳細期刊論文著作請參閱許教授個人網頁。