Açımlayıcı Faktör Analizinde Faktör Sayısı Belirleme Yöntemlerinin Çeşitli Koşullar Altında Karşılaştırılması

Abstract

In this study, it was aimed to compare the factor retention methods (MAP, MAP4, Hull, EGATMFG and Factor Forest) in terms of convergence rate, percent correct and relative bias value. In this Monte Carlo simulation study, simulation conditions were determined as sample size (200, 500, 1000), number of categories of item scores (3, 5 and 7), test length (8 and 16 items), measurement model (unidimensional, orthogonal two-factors and oblique two-factors), distribution of item scores (right-skewed, normal, left-skewed) and average factor loading (0.40, 0.60 and 0.80). According to the fully crossed simulation design, 1000 replications were performed for each of the 486 simulation conditions. As a result of the research, it was determined that none of the methods have convergence problem. It was determined that the increase in sample size and average factor loading had a positive effect on percent correct and relative bias values. Differences were observed between the methods in identifying unidimensional and two-factors. The conditions under which the methods predicted with high accuracy and less bias differed from each other. This situation pointed out the importance of using more than one factor retention method and examining the compatibility of the methods. There is no single method that works accurately and unbiased under all conditions, therefore methods that work well for different conditions have been discussed.