Proposição de dois novos métodos para análise de componentes principais
Year: 2020
Type: dissertation
Abstract: Principal component analysis (PCA) is a multivariate method widely used, mainly because of \nits ability to synthesize in a few latent variables, known as principal components, a large proportion \nof the total variance of all original variables. However, PCA suffers from the fact that \neach principal component is the linear combination of a very large number of original variables, \nwhich often causes difficulties in interpreting the results. One of the ways adopted to overcome \nthis difficulty is to observe the loadings that accompany each variable and ignore those whose \nvalues are small. The component thus obtained becomes the linear combination involving the \nremaining variables. Although this practice is widely used, this procedure is potentially misleading \nas it is based on subjectivity. Sparse principal component analysis (SPCA) has emerged \nas a method that can be applied to improve this disadvantage of PCA. Being a subject of intense \nresearch for over a decade, the SPCA method proposed by Zou, Hastie and Tibshirani \nin 2006 modifies the original formulation of the PCA by treating it as a regression problem by \nintroducing the LASSO penalty, acronym for Least Absolute Shrinkage and Selection Operator, \nwhich is useful for inducing sparse (null loadings) in the principal components. Because of the \nabove, two new methods are proposed in order to facilitate the interpretation of results in the \nPCA, mainly for scenarios in which the problem under investigation has a very large number \nof variables. The proposed methods were called Sparse Group for Principal Component Analysis \n(SGPCA) and Pairwise Absolute Clustering and Sparsity for Principal Component Analysis \n(PACSPCA). The SGPCA and PACSPCA methods are based on the Octogonal Shrinkage and \nClustering Algorithm for Regression (OSCAR) and Pairwise Absolute Clustering and Sparsity \n(PACS) regression methods, respectively. The two new methods proposed, in addition to also \ninducing the sparsity in the components such as the SPCA method, also can group variables \nusing the correlation between them by the equality of their loadings. As an illustration, the \nproposed SGPCA and PACSPCA methods were applied to real and simulated data, aiming to \nelucidate some of their characteristics.
Author: Reis, Carlos José dos
Language: English
Subfield: Statistics and Probability
Field: Mathematics
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W6995848557
