21 Pages Posted: 23 Nov 2013
Date Written: November 21, 2013
We propose a new way to conduct multiple hypothesis testing in economics research. Our framework allows for correlation among tests and incomplete data, both of which are prevalent in economic meta-analysis. Our simulations show that that our method is able to produce the correct p-value cutoff that controls the overall rate of false discoveries at a prespecified level of significance. The single hypothesis test, as used by most researchers, leads to too many false discoveries and should be avoided.
Keywords: Multiple tests, Correlation, Incomplete data, Meta-analysis, Data mining, False discoveries, Type I error, Type II error
JEL Classification: C12, C13, C21, C22, C52, G12, O11, B41
Suggested Citation: Suggested Citation
By Andrew Ang