Published
Comparison of Correction Factors and Sample Size Required to Test the Equality of the Smallest Eigenvalues in Principal Component Analysis
Comparación de los factores de correción y tamaños de muestra requeridos para probar la igualdad de los valores propios más pequeños en el análisis de componentes principales
DOI:
https://doi.org/10.15446/rce.v44n1.83987Keywords:
Chi-square distribution, Likelihood ratio test, Power comparisons, Principal components analysis, Sphericity test (en)Análisis de componentes principales, Comparación de potencias, Distribución Chi-cuadrado, Prueba de esfericidad, Prueba de razón de verosimilitud (es)
Downloads
In the inferential process of Principal Component Analysis (PCA), one of the main challenges for researchers is establishing the correct number of components to represent the sample. For that purpose, heuristic and statistical strategies have been proposed. One statistical approach consists in testing the hypothesis of the equality of the smallest eigenvalues in the covariance or correlation matrix using a Likelihood-Ratio Test (LRT) that follows a χ2 limit distribution. Different correction factors have been proposed to improve the approximation of the sampling distribution of the statistic. We use simulation to study the significance level and power of the test under the use of these different factors and analyze the sample size required for an dequate approximation. The results indicate that for covariance matrix, the factor proposed by Bartlett offers the best balance between the objectives of low probability of Type I Error and high Power.
If the correlation matrix is used, the factors W ∗
and cχ2
are the most
recommended. Empirically, we can observe that most factors require sample sizes 10 or 20
times the number of variables if covariance or correlation
matrices, respectively, are implemented.
Dentro del proceso inferencial del Análisis de Componentes Principales (PCA) uno de los interrogantes principales de los investigadores es sobre el número correcto de componentes para representar la muestra. Para este fin se han propuesto estrategias heurísticas y estadísticas. Un enfoque estadístico consiste en probar la hipótesis sobre la igualdad de los valores propios más pequeños de la matriz de covarianza o correlación a través de una prueba de razón de verosimilitud (LRT) que sigue una distribución límite χ2 . Diferentes factores de corrección han sido propuestos para mejorar la aproximación de la distribución muestral del estadístico. En este trabajo utilizamos simulación para estudiar el nivel de significancia y la potencia de la prueba bajo el uso de estos diferentes factores, así como una revisión del tamaño de muestra requerido para una adecuada aproximación. Los resultados para la matriz de covarianza indican que el factor propuesto por Bartlett ofrece el mejor equilibrio entre los objetivos de baja probabilidad de Error Tipo I y alta potencia. En caso de la matriz de correlación, los factores W ∗ y cχ2 son los
B d
más recomendados. Empíricamente se observa que la mayoría de los factores requieren tamaños de
muestra 10 y 20 veces mayores al número de variables
en caso de la matriz de covarianza o de correlación respectivamente.
References
Anderson, T. (1963), ‘Asymptotic theory for principal component analysis’, The Annals of Mathematical Statistics 34(1), 122–148. DOI: https://doi.org/10.1214/aoms/1177704248
Arteaga, F. & Ferrer, A. (2010), ‘How to simulate normal data sets with the desired correlation structure’, Chemometrics and Intelligent Laboratory Systems 101, 38–42. DOI: https://doi.org/10.1016/j.chemolab.2009.12.003
Bartlett, M. (1951), ‘The effect of standardization on a χ2 approximation in factor analysis’, Biometrika 38(3/4), 337–344. DOI: https://doi.org/10.1093/biomet/38.3-4.337
Bartlett, M. (1954), ‘A note on the multiplying factors for various χ2 approximations’, Journal of the Royal Statistical Society. Series B (Methodological) 16(2), 296–298. DOI: https://doi.org/10.1111/j.2517-6161.1954.tb00174.x
Björklund, M. (2019), ‘Be careful with your principal components’, Evolution 73(10), 2151–2158. DOI: https://doi.org/10.1111/evo.13835
Box, G. E. P. (1949), ‘A general distribution theory for a class of likelihood criteria’, Biometrika 36(3/4), 317–346. DOI: https://doi.org/10.1093/biomet/36.3-4.317
Chakraborty, L., Rus, H., Henstra, D., Thistlethwaite, J. & Scott, D. (2020), ‘A place-based socioeconomic status index: Measuring social vulnerability to flood hazards in the context of environmental justice’, International Journal of Disaster Risk Reduction 43. DOI: https://doi.org/10.1016/j.ijdrr.2019.101394
Ferré, L. (1995), ‘Selection of components in principal component analysis: a comparison of methods’, Computational Statistics & Data Analysis 19, 669–689. DOI: https://doi.org/10.1016/0167-9473(94)00020-J
Friedman, S. (1981), ‘Interpreting the first eigenvalue of a correlation matrix’, Educational and Psychological Measurement 41, 11–21. DOI: https://doi.org/10.1177/001316448104100102
Fujikoshi, Y., Yamada, T., Watanabe, D. & Sugiyama, T. (2007), ‘Asymptotic distribution of the LR statistic for equality of the smallest eigenvalues in high- dimensional principal component analyisis’, Journal of Multivariate Analysis 98, 2002–2008. DOI: https://doi.org/10.1016/j.jmva.2006.10.006
Jackson, D. (1993), ‘Stopping rules in principal components analysis: a comparison of heuristical and statistical approaches’, Ecological Society of America 74(8), 2204–2214. DOI: https://doi.org/10.2307/1939574
Jackson, J. E. (1991), A User’s Guide To Principal Components, John Wiley & Sons, Inc. DOI: https://doi.org/10.1002/0471725331
Jolliffe, I. (2002), Principal Component Analysis, 2 edn, Springer.
Knapp, T. R. & Swoyer, V. H. (1967), ‘Some empirical results concerning the power of Bartlett’ s Test of the significance of a correlation matrix’, American Educational Research Association 4(1), 13–17. DOI: https://doi.org/10.3102/00028312004001013
Krazanowski, W. J. (1988), Principles of Multivariate Analysis, A User’s Perspective, Oxford Statistical Science.
Lawley, D. (1956), ‘Test of significance for latent roots of covariance and correlations’, Biometrika 43(1/2), 128–136. DOI: https://doi.org/10.1093/biomet/43.1-2.128
Mardia, K., Kent, J. & Bibby, J. (1979), Multivariate Analysis, 6 edn, Academic Press, San Diego.
Maté, C. G. (2011), ‘A multivariate analysis approach to forecasts combination. application to foreign exchange (FX) markets’, Revista Colombiana de Estadistica 34(2), 347–375.
Peres-Neto, P. R., Jackson, D. A. & Somers, K. M. (2005), ‘How many principal components? stopping rules for determining the number of non-trivial axes revisited’, Computational Statistics and Data Analysis 49(4), 974–997. DOI: https://doi.org/10.1016/j.csda.2004.06.015
R Core Team (2019), R: A Language and Environment for Statistical Computing, R Foundation for Statistical Computing, Vienna, Austria. https://www.R- project.org/
Şahan, C., Baydur, H. & Demiral, Y. (2018), ‘A novel version of copenhagen psychosocial questionnaire-3: Turkish validation study’, Archives of Environmental & Occupational Health 74(6), 297–309. DOI: https://doi.org/10.1080/19338244.2018.1538095
Schott, J. R. (1988), ‘Testing the equality of the smallest latent roots of a correlation matrix’, Biometrika 75(4), 794–796. DOI: https://doi.org/10.1093/biomet/75.4.794
Schott, J. R. (2006), ‘A high-dimensional test for the equality of the smallest eigenvalues of a covariance matrix’, Journal of Multivariate Analysis 97, 827–843. DOI: https://doi.org/10.1016/j.jmva.2005.05.003
Schott, J. R. (2012), ‘An Approximation for the Test of the Equality of the Smallest Eigenvalues of a Covariance Matrix’, Communications in Statistics-Theory and Methods 41, 4439–4443. DOI: https://doi.org/10.1080/03610926.2011.574219
Watanabe, D., Okada, S., Fujikoshi, Y. & Sugiyama, T. (2008), ‘Large sample approximations for LR statistic for equality of the smallest eigenvalues of a covariance matrix under elliptical population’, Computational Statistics & Data Analysis 52, 2714–2724. DOI: https://doi.org/10.1016/j.csda.2007.09.028
Waternaux, C. (1984), ‘Principal components in the nonnormal case: the test of equality of Q roots’, Journal of Multivariate Analysis 14, 323–335. DOI: https://doi.org/10.1016/0047-259X(84)90037-X
How to Cite
APA
ACM
ACS
ABNT
Chicago
Harvard
IEEE
MLA
Turabian
Vancouver
Download Citation
CrossRef Cited-by
1. Rafael Rodrigues de Souza, Alberto Cargnelutti Filho, Marcos Toebe, Karina Chertok Bittencourt. (2023). Sample size and genetic divergence: a principal component analysis for soybean traits. European Journal of Agronomy, 149, p.126903. https://doi.org/10.1016/j.eja.2023.126903.
2. Alberto Cargnelutti Filho, Marcos Toebe. (2021). Sample size for principal component analysis in corn. Pesquisa Agropecuária Brasileira, 56 https://doi.org/10.1590/s1678-3921.pab2021.v56.02510.
Dimensions
PlumX
Article abstract page views
Downloads
License
Copyright (c) 2021 Revista Colombiana de Estadística

This work is licensed under a Creative Commons Attribution 4.0 International License.
- Authors retain copyright and grant the journal right of first publication with the work simultaneously licensed under a Creative Commons Attribution License that allows others to share the work with an acknowledgement of the work's authorship and initial publication in this journal.
- Authors are able to enter into separate, additional contractual arrangements for the non-exclusive distribution of the journal's published version of the work (e.g., post it to an institutional repository or publish it in a book), with an acknowledgement of its initial publication in this journal.
- Authors are permitted and encouraged to post their work online (e.g., in institutional repositories or on their website) prior to and during the submission process, as it can lead to productive exchanges, as well as earlier and greater citation of published work (See The Effect of Open Access).