International Journal of Mathematics and Statistics Invention (IJMSI) E-ISSN: 2321 – 4767 || P-ISSN: 2321 - 4759 www.ijmsi.org || Volume 2 || Issue 2 || February - 2014 || PP-33-49 www.ijmsi.org 33 | P a g e Hierarchical Component Using Reflective-Formative Measurement Model In Partial Least Square Structural Equation Modeling (Pls-Sem) Wan Mohamad Asyraf Bin Wan Afthanorhan Department of Mathematics, Faculty of Science and Technology, Universiti Malaysia Terengganu, 21030 Kuala Terengganu, Malaysia ABSTRACT: In recent years, partial least square structural equation modeling has been enjoyed popularly since the various package for partial least square established. Besides, this method can be known as the the next second generation modeling or soft modeling that can be a great helpful among the researchers and practitioners to accomplish their objective research. In this paper also intend to modeling the second higher order construct (Hierachical Component) as the advance in partial least structural equation modeling (PLS- SEM) using smartpls which is the newest package. In this application of this method, we can create a higher order construct, in particular, the reseracher should empahsize for many aspect in order to ensure this model is more relevance and significant. Thus, the application using reflective-formative should be carry out in order to obtain the best model. In some instance, the author present the guideline to conduct this analysis with a real example so that the researchers outside will be more understanding and enjoyed for this new application. KEYWORDS: Partial Least Square Structural Equation Modeling, Hierarchical Component Model, Second Order Construct, Reflective-Formative Model, SmartPls I. INTRODUCTION PLS-SEM has been established for a long time ago by Wold (1982), however, this method is not popular as covariance based structural equation modeling (CB-SEM) in which focuses on goodness of fitness to minimize the covariance matrix and estimation matrix (Hair, 2010) earlier 1980. However, CB-SEM has a lot weakness since the reserchers should be ensure the model has achieved requirement before subsequent analysis in the structural model. In this case, the researcher has been spent time to focus on goodness of fit rather than estimation or prediction. Therefore, the introduction to PLS-SEM is returned now with a great helpful and more user friendly to curb the problem of researchers nowadays. In the accordance with Hair (2010) discover the PLS-SEM is aimed to maximize the explained variance of the endogenous construct (square multiple correlation, R 2 ) of the endogenous latent construct (dependent). This application is performed nonparametric analysis in which does not rely on distributional assumption (Chin, 1998). Thus, this method is appropriate for those who have insufficient data, time and others. However, PLS-SEM is does not assume data to be normal even it appropriate for nonparametric. Thus, the bootstrap in smartpls is used to resampling the data until the data meet the result. According to Byrne (2010), bootstrap is an aid for nonparametric data in structural equation modeling. Hair (2010) listed several advantages for those who apply PLS-SEM:  Normality of data distribution not assumend normality  Can be used with fewer indicator (manifest variable)  Models can be include a larger number of indcator variable  Preffered alternative with formative construct  Assumes all measured variance (including error) is useful for explanation/prediction of causal relationship The result obtained in t-distribution against CB-SEM since this method is performed nonparametric analysis. In addition, the researcher does not have difficult to apply the formative construct in PLS-SEM. Formative construct in CB-SEM is much complicated than PLS-SEM and, of course, PLS-SEM ease the researchers to perform their analysis regarding on their objective research.