PhilSci Archive

Inflated Effect Sizes, Underpowered Tests and the Severity Measure of Evidence

Rochefort-Maranda, Guillaume (2017) Inflated Effect Sizes, Underpowered Tests and the Severity Measure of Evidence. [Preprint]

[img]
Preview
Text
manusc_infl.pdf

Download (536kB) | Preview

Abstract

The severity score is particularly high for hypotheses that are substantially different from the null-hypothesis when a significant result is obtained by using an underpowered test. This means that such hypotheses are very well supported by the evidence according to that measure.
However, it is now well documented that significant tests with low power display inflated effect sizes. They systematically show departures from the null hypothesis H0 that are much greater than they really are. This is problematic in research contexts where the differences between H0 and H1 is particularly small and where the sample size is also small.
In this paper I argue that the severity score is an inadequate measure of evidence and that it should be rejected. The reason is that it is sensitive to the inflated effect sizes provided by underpowered significant tests: inflated effect sizes also inflate severity scores.


Export/Citation: EndNote | BibTeX | Dublin Core | ASCII/Text Citation (Chicago) | HTML Citation | OpenURL
Social Networking:
Share |

Item Type: Preprint
Creators:
CreatorsEmailORCID
Rochefort-Maranda, Guillaumeguillaumemaranda@hotmail.com
Keywords: error-statistical philosophy, inflated effect sizes, severity score, underpowered tests
Subjects: Specific Sciences > Probability/Statistics
Depositing User: Dr. Guillaume Rochefort-Maranda
Date Deposited: 14 Apr 2017 12:33
Last Modified: 14 Apr 2017 12:33
Item ID: 12986
Subjects: Specific Sciences > Probability/Statistics
Date: 13 April 2017
URI: http://philsci-archive.pitt.edu/id/eprint/12986

Monthly Views for the past 3 years

Monthly Downloads for the past 3 years

Plum Analytics

Actions (login required)

View Item View Item