How Do Partial Least Square Structural Equation Modeling and Covariance-based Structural Equation Modeling Vary from One Another?
Covariance-based structural equation modeling (CB-SEM) and partial least squares structural equation modeling (PLS-SEM) are two methods for estimating structural equation models.
CB-SEM is a method for estimating and testing the relationships between observed variables and latent constructs in a model. It is based on the assumption that the observed variables are measured with error and that the relationships between the observed variables and latent constructs can be represented by a set of regression equations. CB-SEM estimates the model parameters by maximizing the likelihood of the data given the model.
PLS-SEM is a method for estimating and testing the relationships between observed variables and latent constructs in a model. It is based on the assumption that the observed variables are correlated with each other and with the latent constructs, and that the relationships between the observed variables and latent constructs can be represented by a set of regression equations. PLS-SEM estimates the model parameters by minimizing the sum of squared residuals between the observed variables and their predicted values.
There are several key differences between CB-SEM and PLS-SEM:
Assumptions: CB-SEM assumes that the observed variables are measured with error, whereas PLS-SEM assumes that the observed variables are correlated with each other and with the latent constructs.
Estimation method: CB-SEM estimates the model parameters by maximizing the likelihood of the data given the model, while PLS-SEM estimates the model parameters by minimizing the sum of squared residuals between the observed variables and their predicted values.
Sample size: CB-SEM requires a larger sample size than PLS-SEM to achieve good model fit.
Model fit: CB-SEM is more sensitive to model misspecification and may produce biased parameter estimates when the model is misspecified, whereas PLS-SEM is more robust to model misspecification and may produce unbiased estimates even when the model is misspecified.
Applicability: CB-SEM is suitable for testing relationships between observed variables and latent constructs, whereas PLS-SEM is more suitable for predicting the values of observed variables from latent constructs.