Linear Latent Structure Analysis: from Foundations to Algorithms and Applications

Abstract

A new statistical technique for constructing linear latent structure (LLS) models from available data, supported by well established theoretical results and an efficient algorithm, is presented. The method reduces the problem of estimating LLS model parameters to a sequence of linear algebra problems. This assures a low computational complexity and an ability to handle large scale data that involve thousands of variables. An overall computational scheme and all its components are discussed in detail. Simulation experiments demonstrate the excellent performance of the algorithm in reconstructing model parameters. Step-by-step analysis of a demographic survey is presented as an example. The technique is useful for the analysis of high-dimensional categorical data (e.g., demographic surveys, gene expression data) where the detection, evaluation, and interpretation of a underlying latent structure are required.