Learning Confirmatory Patterns in Exploratory Factor Analysis Using ICOMP and Genetic Algorithm

The dissertation intends to develop a new approach to the identification of the best factor pattern structure. This new approach is a multivariate regression analysis where factor scores are regressed on original variables. The dissertation shows the versatility of information model selection criteria, Bozdogan's ICOMP- type criteria in particular, in two types of modeling problems: determining the number of factors in factor analysis and working as the fitness function for Genetic Algorithm.