Confirmatory Factor Analysis (CFA) Model for Testing Normality with the Weight Least Square (WLS) Estimation Method

Abstract

To complete the analysis problem of the data which form ordinal scale, we used Structural Equation Modeling (SEM) or Lisrel. This research aims to evaluate the normality and covariant matrix estimation from the ordinal data which unknown the spreading form in each sample 100, 150, 200, and 300 by using Weighted Least Square (WLS). Ordinal data appeared from the spreading of lisrel aid and used matrix covariant aid. Path of diagram also used to test of the suitable model with WLS. The normally graphic of each sample test are the Kolmogorov-Smirnov test, Anderson-Darling test, and Ryan-Joiner test. Determination of model quality could be used based on three groups of the model, they are absolute model:  and RMSEA, incremental model: GFI and AGFI, and parsimony model: PGFI. This research found the normality by using the WLS method in samples 100, 150, 200, and 300 by using Kolmogorov-Smirnov test, Anderson-Darling test, Ryan-Joiner test. According to the result, it could be found that for samples 100, 150, 200, and 300, the spread of the data was normal and the best test used the Kolmogorov-Smirnov test. This result shows that if the sample has a larger size, it means the sample has the best value in the test performed.