Detection of False Investment Strategies through FWER and FDR (Seminar Slides)

Abstract

Financial systems rarely allow experimentation. For example, we cannot reproduce the flash crash of 2010 while controlling for environmental conditions. As a result, much financial research relies on the statistical analysis of finite (historical) datasets, where: (a) Time series datasets are limited, and (b) The investment universe is limited.

The implication is that a large number of hypotheses are tested on the same observations. In the context of asset management, this situation leads to false investment strategies and losses, particularly among quantitative funds.

This seminar explains how to detect false investment strategies by controlling for the familywise error rate (FWER) and the false discovery rate (FDR) of an organization. It is part of Cornell University's ORIE 5256 course.

Keywords: Multiple testing, p-values, p-hacking, selection bias, familywise error rate, false discovery rate

JEL Classification: G0, G1, G2, G15, G24, E44