| Summary: | Unweighted least squares (ULS) is used in confirmatory research to handle nonnormal data. While ULS considers measurement errors in observed variables, it frequently results in inaccurate solutions, such as negative estimates or boundary values for unique variances. In confirmatory factor analysis, unique variance is portrayed as a disturbance. This includes consistent variation in the item that discloses unknown latent causes and random error generated from measurement error or unreliability. Consequently, this may lead to bias in the estimate of indicator loadings. To resolve this issue, a regularization parameter is added to each element in the sample variance-covariance matrix of the ULS estimator. Numerous values for the regularization parameter (λ) are tested, and the optimal value is determined based on the model's performance. Specifically, the regularization parameter associated with the smallest RMSEA is selected for each model. This method aims to optimize the balance between bias and variance in the estimation of indicator loadings. For this study, family well-being dataset obtained from the National Population and Family Development Board (NPFDB) was used. The analysis was carried out using R Programming Environment using lavaan package. The results of this study demonstrated that the regularized ULS is able to enhance indicator loading estimates in attaining unidimensional validity for family well-being dimensions. © 2024 Author(s).
|
|---|