An evaluation of the Nest Eigenvalue Sufficiency Test (NEST) [r-libre/3023]

Caron, Pier-Olivier (juill. 2023). An evaluation of the Nest Eigenvalue Sufficiency Test (NEST). Communication présentée au 10e European Congress of Methodology de l'European Association of Methodology, Gand.

Fichier(s) associé(s) à ce document :

PDF - caron2023_EAM.pdf
Contenu du fichier : Diaporama

Télécharger

Catégorie de document :	Communications à des congrès/colloques et conférences (non publiées)
Évaluation par un comité de lecture :	Oui
Étape de publication :	Non publié
Résumé :	Determining the number of factors to retain in an exploratory factor analysis is an open methodological problem still after 75 years of research. A plethora of techniques exist to address this challenge. One of the most promising techniques is the Next Eigenvalue Sequence Test (NEST; Achim, 2017), which shows excellent performance (Achim, 2020; Brandenburg & Papenberg, 2022), but has not, however, been systematically compared to well-established competitors (see, Auerswald & Moshagen, 2019). The present study thus proposes a simulation with synthetic factor structures to compare NEST, parallel analysis, minimum average partial correlation, χ^2 sequential test, Hull method, and the empirical Kaiser criterion. The structures are based on 24 variables containing 1 to 8 factors, with loadings ranging from .40 to .80, inter-factor correlations from .00 to .30, on three sample sizes, 120, 240 and 480. In total, 360 scenarios are tested 1000 times. Performance is evaluated in terms of accuracy (correct identification of dimensionality), bias (tendency to over- or underestimate dimensionality) and variability (magnitude of prediction error). The results show that NEST outperforms these competitors. While most techniques do well in easy scenarios, NEST particularly stands out in difficult ones. Scenarios in which all methods fail are discussed. Some limitations of NEST are adressed. Finally, a new R package, named RNest, which implements NEST is promoted.
Adresse de la version officielle :	https://eam2023.ugent.be/
Déposant:	Caron, Pier-Olivier
Responsable :	Pier-Olivier Caron
Dépôt :	02 août 2023 15:36
Dernière modification :	02 août 2023 15:36

Actions (connexion requise)

RÉVISER