Classical principal component analysis

Publication date 18 November 2001

data matrices principal components analysis

Factor analysis and principal components; correspondence analysis (62H25)

Here, the classical PCA is briefly sketched. Classical principal components analysis starts with \(n\) data points \(x_1,\dots,x_n\) in a \(p\)-dimensional Euclidean space \(\mathbb{R}^p\). In this case, \(x_k=(x_{k1},\dots,x_{kp})'\) (a column vector) describes the properties of the object \(k\in\Omega=\{1,\dots,n\}\) in terms of the values \(x_{kj}=Y_j(k)\) which were observed for \(p\) quantitative variables \(Y_j\) with domains \({\mathcal Y}_j\) in \(\mathbb{R}\). These data are usually compiled in a classical data matrix \(X=(x_{kj})_{n\times p}\). For low dimensions \(p=1,2,3\), the data points can be exactly and directly visualized on the real line \(\mathbb{R}\), as a two-dimensional display in \(\mathbb{R}^2\), or as a \(3D\)-picture, respectively, which leads typically to a direct substantial interpretation of the data. However, for higher dimensions we are faced with the general problem of how to visualize high-dimensional data points \(x_1,\dots,x_n\in \mathbb{R}^p\) by a low-dimensional configuration of points \(z_1,\dots,z_n\in \mathbb{R}^s\) with dimensions \(s=2\) or \(s=3\). Classical principal components analysis resolves this problem. A proof of the corresponding construction, various interpretations in terms of ``maximum variance, and many illustrative examples can be found in the current textbooks on multivariate data analysis [see, for example, \textit{N.C. Giri}, Multivariate statistical analysis. (1996; Zbl 0846.62039)]. For the entire collection see [Zbl 1039.62501].

Recommendations

Cited in

(1)

Projection of a correlation matrix onto the unit sphere in \(\mathbb{R}^3\)

This page was built for publication: Classical principal component analysis

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2756640)