Non-Graphical Solutions for Cattell’s Scree Test

By Gilles Raîche, Theodore A. Walls, David Magis, Martin Riopel, J. Blais. 2013

326 citations recorded by the source corpus.

Most of the strategies that have been proposed to determine the number of components that account for the most variation in a principal components analysis of a correlation matrix rely on the analysis of the eigenvalues and on numerical solutions. The Cattell's scree test is a graphical strategy with a nonnumerical solution to determine the number of components to retain. Like Kaiser's rule, this test is one of the most frequently used strategies for determining the number of components to retain. However, the graphical nature of the scree test does not definitively establish the number of components to retain. To circumvent this issue, some numerical solutions are proposed, one in the spirit of Cattell's work and dealing with the scree part of the eigenvalues plot, and one focusing on the elbow part of this plot. A simulation study compares the efficiency of these solutions to those of other previously proposed methods. Extensions to factor analysis are possible and may be particularly useful with many low-dimensional components. Several strategies have been proposed to determine the num- ber of components that account for the most variation in a principal components analysis of a correlation matrix. Most of these rely on the analysis of the eigenvalues of the corre- lation matrix and on numerical solutions. For example, Kaiser's eigenvalue greater than one rule (Guttman, 1954; Kaiser, 1960), parallel analysis (Buja & Eyuboglu, 1992; Horn, 1965; Hoyle & Duvall, 2004), or hypothesis signifi- cance tests, like Bartlett's test (1950), make use of numerical criteria for comparison or statistical significance criteria. Independently of these numerical solutions, Cattell (1966) proposed the scree test, a graphical strategy to determine the number of components to retain. Along with the Kaiser's rule, the scree test is probably the most used strategy and it is included in almost all statistical software dealing with principal components analysis. Unfortunately, it is generally recognized that the graphical nature of the Cattell's scree test does not enable clear decision-making about the number of components to retain. The previously proposed non-graphical solutions for

Subfields: Food Science, Statistics and Probability

View Paper: https://www.semanticscholar.org/paper/e8756ee4fe3fe3db4cef1bee669c19b8ab8d44f7

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Food Science

In Statistics and Probability

Citations in the corpus

Measuring procrastination at work and its associated workplace aspects (2016, unknown): “Four eigenvalues exceeded 1.00, with values of 5.85, 1.56, 1.23 and 1.07, respectively, showing that the ﬁ nal solution could potentially have one, two, three or four factors (Raîche, Walls, Magis, Riopel, & Blais, 2013).”
One Size Doesn’t Fit All: Using Factor Analysis to Gather Validity Evidence When Using Surveys in Your Research (2019, methodology): “In addition, several statistics that mathematically analyze the shape of the scree plot have been developed in an effort to provide a nonvisual method of determining the number of factors (Ruscio and Roche, 2012; Raiche et al., 2013).”
Investigating the performance of Exploratory Graph analysis and traditional techniques to identify the number of latent factors: a simulation and tutorial (2018, methodology): “The scree test optimal coordinate (OC) and acceleration factor (AF) methods (Raiche, Walls, Magis, Riopel, & Blais, 2013) constitute two nongraphical solutions to Cattell’s scree test (Cattell, 1966).”