PARTIAL LEAST SQUARE PATH MODELING (PLS PM) ANALYSIS WITH RESPONSE BASED UNIT SEGMENTATION IN PARTIAL LEAST SQUARE (REBUS PLS)

• Main Author: Ayudhia Sasmito, Karina
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/73518
• Institution: Institut Teknologi Bandung

Partial Least Square Path Modeling (PLS PM) is an effective analysis method, because it can be used on any type of data scale and more flexible assumption requirements. In estimation using PLS PM, the basic principle used in the data is assumed to be homogeneous in all units of the observation set. This implies that the treatment of data ignores the existence of a group structure and all units are assumed to be well represented by a single estimation model in the entire sample called the global model. If the assumed homogeneity in the structural model does not hold in the data model, then incorrect/incomplete conclusions will be obtained. So that a solution is needed for the existence of diversity in the data or heterogeneity. To overcome this heterogeneity, using Response Based Unit Segmentation in Partial Least Square (REBUS PLS) is an iteration algorithm to overcome heterogeneity in measurement and structural equation models. This study analyzes the PLS PM model with REBUS PLS in the structural model of the effect of communication and transparency relationships on the quality of public services, a case study of the Bugis Village Office. There are three latent variables, namely communication, transparency and public service quality. In addition, the exogenous variable of communication consists of five indicators, the endogenous variable of transparency consists of four indicators and the endogenous variable of public service quality consists of five indicators. The results of applying REBUS PLS to the PLS PM model of the effect of communication and transparency on the quality of public services show that observation units can be grouped into two and three segments. If two segments are formed, the ????2 and GoF values of each local model have better values than the ????2 and GoF of the global model. If there are three segments, then one of the local models shows lower ????2and GoF values than the global model. The formation of these segments on the unit of observation indicates that there is heterogeneity in the global model.