Feature Space Modelling

Abstract

It is rather remarkable that the stochastic linear state space model and Jöreskog’s general linear structural model, extensively used in social sciences, are — to a certain extent — equivalent in structure, see e.g. Jöreskog (1977). JÖreskog’s model is, however, static and structural, whereas the stochastic linear state space model is (time) dynamic, but not structural. By structural we mean that there exists a relationship between output variables (or output factors). In section 3.2 a general model is formulated, called the linear structural feature space (LSF) model, which essentially features the main characteristics of the two models. It is a linear, dynamic, structural model, which consists of two equations: a structural linear dynamic process equation in the feature space and a linear output or measurement equation. The reduced form version of the (LSF) model, denoted by (LRF) model can be seen as a linear stochastic state space model with stochastic inputs, see for details Otter (1981a).KeywordsWhite NoiseSingular Value DecompositionCanonical CorrelationModel ReductionCanonical VariableThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.