STUDY OF SEMIVARIOGRAM DISTRIBUTION AND ITS APPLICATION ON POVERTY DATA OF BANDUNG CITY

• Main Author: KAMALITA (NIM:10114096), YOLA
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/31759
• Institution: Institut Teknologi Bandung

Sub-districts in Bandung City have various poverty rate. Spatial relationships between sub-district locations can be illustrated by the semivariogram the variance of the difference in poverty rates from a pair of locations. In the case of poverty, there are two sub-districts with poverty rates detected as outliers. Therefore, Dowd's experimental semivariogram approach is used, because of the robustness of the outliers. One of the information needed to estimate the parameters of the semivariogram model using Maximum Likelihood (ML) method is semivariogram distribution. The analytic distribution of semivariogram is chi-square, whereas the simulation result shows that the log-normal distribution is always significant enough to approximate the semivariogram distribution. The parameters of semivariogram model (partial sill C, nugget effect C_0 and range a) will be estimated with Maximum Likelihood Iterative (MLI) Modification. Values a and C will be estimated simultaneously with the specified value a will minimize Sum of Squared Errors (SSE). Then, Jacknife Kriging is applied to validate the semivariogram model. The Jacknife Kriging process provides an estimated value of all sub-districts by estimating the poverty rate of an erased subdistrict. The smallest Kriging SSE value determines the best semivariogram model. In this case, the Gauss model is chosen to be the best model because it has the smallest Kriging SSE 25,5658. The estimated value of a from the Gauss model is 1.8 kilometers. This means that in a radius of 1.8 kilometers the poverty rate among locations will be similar. Thus, it is necessary to watch out for the influence of the sub-district with the high poverty rate toward the surrounding area within the radius.