Abstract
The necessity of performing exploratory factor analysis (EFA) for predetermined constructs has been debated among researchers; some argue that performing EFA is essential before confirmatory factor analysis (CFA), and others contend that EFA is unnecessary when the items under construction have been predetermined. This study seeks to contribute to the ongoing discussion by examining whether it is necessary to perform EFA prior to conducting CFA when a theoretical model has already been established. To mimic the real-life scenarios, population data of size, n=500 with predetermined relationships between items and constructs was generated by using the Monte-Carlo Markov Chain method via the “MASS”, “mvrnorm”, and “psych” packages in R programming. The data generated for the population dataset were prespecified factor loadings for items under exogenous and endogenous constructs were set to 0.6 and 0.7, respectively. Next, samples of varying sizes (n=50, 100, and 300) were randomly selected from the generated population data. The results indicate that EFA yields unsatisfactory outcomes across all sample sizes (n=50, 100, and 300), as it failed to adequately discern items under predetermined constructs, as in the population dataset. Therefore, it is concluded that EFA is unsuitable for studies with predetermined constructs, especially when the sample size is less than 300.
Keywords: Confirmatory Factor Analysis, Exploratory Factor Analysis, Predetermined Model, Sample Survey.