The classical data situation
categorical variableslogical dependencemissing observationsobservation spacequalitative variablesquantitative variablesrangestochastic dependencetaxonomic dependencetaxonomic variables
This chapter deals with the classical paradigm of data analysis, defines the elementary concepts and notations, explains the classical types of data and illustrates the definitions by examples. Classical data analysis starts with a given number \(n\) of individuals (often termed ``objects) which are collected in a set \(\Omega=\{1,\dots,n\}\) of objects. Properties of each individual are characterized by \(p\) variables \(Y_1,\dots,Y_p\) defined on \(\Omega\). Each variable \(Y_j\) takes its values in a certain set \(\mathcal Y_j\) of possible levels, alternatives or numbers, which is termed its range, domain or observation space, and for each individual \(k\in\Omega\) the variable \(Y_j\) realizes just one single value \(Y_j(k)\) from \(\mathcal Y_j\) which is denoted by \(x_{kj}\). The classical variable \(Y_j\) is defined as a single-valued mapping from \(\Omega\) to \(\mathcal Y_j\) such that \(x_{kj}=Y_j(k)\) is the value observed for the individual \(k\). All these data are compiled in a data matrix \(X=\{x_{kj}\}_{k=\overline{1,n}}^{j=\overline{1,p}}\). Depending on the size of \(\mathcal Y_j\) and on the additional structure imposed on the elements \(x,y,\dots\in\mathcal Y_j\), we can distinguish two major types of variables: quantitative (metric, numerical) and qualitative (nominal, categorical or classificatory, taxonomic) variables. Simultaneously there are considered various measures of proximity \(\delta_j\) between pairs of elements \(x,y\in Y_j\) such that \(\delta_j(x,y)\) measures how ``different or ``dissimilar those elements are considered in the underlying domain. Such proximities will be useful when defining the similarity or dissimilarity of individuals \(k\in\Omega\) on the basis of the observed \(p\) variables. Data analysis is often concerned with dependent variables. In the context of the Symbolic Official Data Analysis System (SODAS) project the following dependencies are distinguished: logical dependence, taxonomic dependence (``mother - daughter variables), and stochastic dependence. For the entire collection see [Zbl 1039.62501].
This page was built for publication: The classical data situation
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2756623)
