Abstract
Comparison of maximum likelihood and unweighted least squares estimation methods in confirmatory factor analysis by Monte Carlo simulation. This article examines the recovery of weak factors in the context of confirmatory factor analysis. Previous research only refers to exploratory factor analysis. The study is done by Monte Carlo simulation with the following conditions: comparison of maximum likelihood (ML) and unweighted least squares (ULS) estimation methods, sample size (100, 300, 500, 1.000 and 2.000) and level of factor weakness (loadings of 0.25, 0.40 and 0.50). Results show that with small sample sizes ML failed to recover the weak factor while ULS succeed in many cases. This advantage is related to the occurrence of Heywood cases. Also the weak factor recovery improves as the level of factor weakness decreases and the number of factors in the model increases.
