We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

econ.EM

Change to browse by:

References & Citations

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Economics > Econometrics

Title: Extended MinP Tests of Multiple Hypotheses

Authors: Zeng-Hua Lu
Abstract: Much empirical research in economics and finance involves simultaneously testing multiple hypotheses. This paper proposes extended MinP (EMinP) tests by expanding the minimand set of the MinP test statistic to include the $p$% -value of a global test such as a likelihood ratio test. We show that, compared with MinP tests, EMinP tests may considerably improve the global power in rejecting the intersection of all individual hypotheses. Compared with closed tests EMinP tests have the computational advantage by sharing the benefit of the stepdown procedure of MinP tests and can have a better global power over the tests used to construct closed tests. Furthermore, we argue that EMinP tests may be viewed as a tool to prevent data snooping when two competing tests that have distinct global powers are exploited. Finally, the proposed tests are applied to an empirical application on testing the effects of exercise.
Comments: 47 pages, 9 figures, 6 tables
Subjects: Econometrics (econ.EM)
Cite as: arXiv:1911.04696 [econ.EM]
  (or arXiv:1911.04696v1 [econ.EM] for this version)

Submission history

From: Zeng-Hua Lu [view email]
[v1] Tue, 12 Nov 2019 06:12:56 GMT (1130kb,D)

Link back to: arXiv, form interface, contact.