Year of Graduation
Comparative Analysis of Factorization Methods in Sociological Research
Applied Methods of Social Analysis of Markets
Author compares situational adequacy of seven factor extraction methods - principal components analysis, principal axis factoring, generalized least squares method, unweighted least squares method, maximum likelihood method, alpha-factoring and image factoring - in different research situations involving six parameters of "ideal" input data for factor analysis restricted by sample size, number and distribution of indicators, model error, size and range of communalities. Factor extraction method selection algorithm was developed by means of literature analysis and empirically tested via statistical experiment with MCMC simulations. Based on experiment results author recommends: a) using principal axis factoring or alpha-factoring if there is no model error and b) if there is a possibility of model error, choosing between maximum likelihood method and generalized least squares method based on sample and communalities size and the distribution of indicators.