How to proceed when both normality and sphericity are violated in the repeated measures ANOVA
Adjusted F-tests have typically been proposed as an alternative to the F-statistic in repeated measures ANOVA. Despite considerable research, it remains unclear how these statistics perform under simultaneous violation of normality and sphericity. Accordingly, our aim here was to conduct a detailed...
| Autores: | , , , , |
|---|---|
| Tipo de recurso: | artículo |
| Estado: | Versión publicada |
| Fecha de publicación: | 2024 |
| País: | España |
| Institución: | Varias* (Consorci de Biblioteques Universitáries de Catalunya, Centre de Serveis Científics i Acadèmics de Catalunya) |
| Repositorio: | Recercat. Dipósit de la Recerca de Catalunya |
| OAI Identifier: | oai:recercat.cat:2445/218943 |
| Acceso en línea: | https://hdl.handle.net/2445/218943 |
| Access Level: | acceso abierto |
| Palabra clave: | Anàlisi de variància Ciències socials Models lineals (Estadística) Analysis of variance Social sciences Linear models (Statistics) |
| Sumario: | Adjusted F-tests have typically been proposed as an alternative to the F-statistic in repeated measures ANOVA. Despite considerable research, it remains unclear how these statistics perform under simultaneous violation of normality and sphericity. Accordingly, our aim here was to conduct a detailed examination of Type I error and power of the F-statistic and the Greenhouse-Geisser (F-GG) and Huynh-Feldt (F-HF) adjustments, manipulating the number of repeated measures (3-6), sample size (10-300), sphericity (epsilon estimator of Greenhouse-Geisser from its lower to upper limit), and distribution shape (slight to extreme deviations from normality). The findings show that the behavior of F-GG and F-HF depends on the degree of violation of both normality, sphericity, and sample size. Overall, we suggest using F-GG under violation of sphericity and slight or moderate deviations from normality in all sample size; with severe deviations from both normality and sphericity F-GG may be used with a sample size larger than 10; and with extreme deviation from both normality and sphericity this statistic may be used with a sample size larger than 30. In the event of discrepant results between F-GG and F-HF, the choice depends on the epsilon value. |
|---|