Have a personal or library account? Click to login
Why Psychologists Should by Default Use Welch’s t-test Instead of Student’s t-test Cover

Why Psychologists Should by Default Use Welch’s t-test Instead of Student’s t-test

Open Access
|Apr 2017

Figures & Tables

irsp-30-82-g1.png
Figure 1

Estimated power of Levene’s test as a function of sample size, SDR and centering parameter.

Table 1

Comparison of t-value and Degrees of Freedom of Welch’s and Student’s t-test.

Equal variancesUnequal variances
Balanced designtWelch = tStudenttWelch = tStudent
dfWelch = dfStudentdfWelch ≠ dfStudent
pWelch = pStudentpWelch ≠ pStudent
Unbalanced designtWelch = tStudenttWelch ≠ tStudent
dfWelch ≠ dfStudentdfWelch ≠ dfStudent
pWelch ≠ pStudentpWelch ≠ pStudent
irsp-30-82-g2.png
Figure 2

P-value distributions for Student’s and Welch’s t-test under the null as a function of SDR, and sample size.

irsp-30-82-g3.png
Figure 3

P-values from Student’s t-test against p-values from Welch’s t-test under the null.

DOI: https://doi.org/10.5334/irsp.82 | Journal eISSN: 2397-8570
Language: English
Published on: Apr 5, 2017
Published by: Ubiquity Press
In partnership with: Paradigm Publishing Services
Publication frequency: 1 issue per year

© 2017 Marie Delacre, Daniël Lakens, Christophe Leys, published by Ubiquity Press
This work is licensed under the Creative Commons Attribution 4.0 License.