Comparison of Rebus PLS and Fimix PLS for Estimating Parameter in Structural Model

Partial Least Square Path Modeling (PLS-PM) is simple method in modeling involved latent variables. It can apply to small of sample size and unspecified distribution of units. In heterogenous units exists, a segmentation approach that is the development of PLS needs to bias minimizing. The popular approaches are Fimix PLS and Rebus PLS. Fimix PLS detects heterogeneity from structural model and more Rebus PLS detects it both structural and measurement model. Both are compared in overlapping units in the segment and variance of residual. The performance was measured by mean bias absolute (MBA) for parameter in structural model. Rebus PLS was given more stable parameter estimator in all conditions. However, Fimix PLS showed the best performance in all conditions.