Learning Factor Patterns in Exploratory Factor Analysis Using the Genetic Algorithm and Information Complexity as the Fitness Function
|
Learning Factor Patterns in Exploratory Factor Analysis Using the Genetic Algorithm and Information Complexity as the Fitness Function
Hongwei Yang, Hamparsum Bozdogan
JPRR Vol 6, No 2 (2011); doi:10.13176/11.335
Hongwei Yang, Hamparsum Bozdogan
Abstract
This paper presents a new, and novel data-adaptive expert approach to determining the best factor pattern structure in exploratory factor analysis (EFA) models using a clever genetic algorithm (GA) hybridized with information theoretic complexity (ICOMP) criterion as the fitness function. These factor pattern structures from EFA model could then be utilized for various inductive inferences for example to study substantive hypotheses of researchers in the light of the data or could be used as data-adaptive prior information in a Bayesian Factor Analysis (BFA) model among other uses.
Numerical applications are shown on a large scale Monte Carlo study and on a real benchmark data set to demonstrate the versatility of ICOMP and AIC-type information criteria as the fitness functions in GA, in two types of modeling problems: (i) Choosing the number of factors in EFA, and (ii) determining the best factor pattern structure in EFA models in one expert system.
Numerical applications are shown on a large scale Monte Carlo study and on a real benchmark data set to demonstrate the versatility of ICOMP and AIC-type information criteria as the fitness functions in GA, in two types of modeling problems: (i) Choosing the number of factors in EFA, and (ii) determining the best factor pattern structure in EFA models in one expert system.