Hi teacher. Please help me with the following: 1.- What is the difference between using the polychoric correlation matrix vs Pearson for an EFA? In social sciences, it is common to work with a Likert scale (1 to 5). 2.- What is your opinion of using the parallel method to explore the number of factors? 3.- Which method do you recommend or have the best performance: MINRES, Ml, PAF? Thank you very much!
1) Polychoric assumes an underlying latent variable but only binary indicators are observed. In Pearson, we assume that the variables that we observe are of direct interest. See Rhemtulla, M., Brosseau-Liard, P. É., & Savalei, V. (2012). When can categorical variables be treated as continuous? A comparison of robust continuous and categorical SEM estimation methods under suboptimal conditions. Psychological Methods, 17(3), 354-373. dx.doi.org/10.1037/a0029315 and the broader IRT literature. 2) I do not now what "parallel method" is so I do not have an opinion on it. 3) Depends on the context. ML is a good default, PAF a fall back, and MINRES is good for teaching but not for anything else.
