| dc.description.abstract | In analyzing with statistics, there are variables that cannot be observed directly (unobserved variables). Unobserved variables can be analyzed by the Structural Equation Modeling (SEM) method. SEM has 2 types of analysis, namely Covariance Based (CB-SEM) and SEM Partial Least Square (SEM-PLS). SEM-PLS has the advantages of unconditional normality and minimum samples. This certainly facilitates research involving unobserved variables as well as can help research that has a small sample. However, due to unconditional normality, the possibility of heterogeneous data is very high. Therefore, segmentation is needed to group data based on the similarity of its characteristics so that better results are obtained. The method that can create this segmentation is Finite Mixture Partial Least Square (FIMIX-PLS). The application of SEM-PLS analysis to poverty structure in Indonesia obtained a positive relationship between: health to education by 0.867 and education to the economy by 0.832. As well as the negative relationship between the human development index to poverty of 0.544 and the economy to poverty of 0.404. The positive influence of the economy on the human development index amounted to 0.884. The suspicion of heterogeneity exists because without the assumption of normal data distribution and data for each province is diverse. So segmentation with FIMIX-PLS is needed. The provincial segmentation formed is 5 segments with a Normed Entropy (EN) value of 0.934. Segment 1 consists of 10 provinces, segment 2 consists of 8 provinces, segment 3 consists of 6 provinces, segment 4 consists of 5 provinces, and segment 5 consists of 3 provinces. | en_US |
