IMPLEMENTASI MODEL LOCALLY COMPENSATED RIDGE-GEOGRAPHICALLY WEIGHTED REGRESSION PADA DATA SPATIAL DENGAN MULTIKOLINEARITAS (Studi Kasus: Stunting Balita di Provinsi Nusa Tenggara Timur)
| Main Authors: | Fadliana, Alfi, Pramoedyo, Henny, Fitriani, Rahma |
|---|---|
| Format: | Article info Journal |
| Bahasa: | eng |
| Terbitan: |
Department of Statistics, Faculty of Science and Mathematics, Universitas Diponegoro
, 2020
|
| Subjects: | |
| Online Access: |
https://ejournal.undip.ac.id/index.php/media_statistika/article/view/23603 |
Daftar Isi:
- NTT Province based on the results of the 2013 Riskesdas and the 2016 and 2017 PSG was recorded as the province with the highest prevalence of stunting nationally. Efforts should be made to formulate policies that are integrated with spatial aspects in order to reduce the prevalence of stunting. The GWR model approach might be the right choice because it can overcome the problem of spatial heterogeneity. However, GWR is not able to accommodate multicollinearity problems that have the potential to occur at the local regression coefficient between predictor variables. Therefore the LCR-GWR model approach is used which is the development of the GWRR model, using locally compensated ridge, ridge bias coefficients for observation region adjusts to the effect of collinearity between predictor variables in each region. The use of LCR-GWR is expected to be able to accommodate the problem of local multicollinearity so that it can be obtained more accurate estimation of model coefficient parameters. From the analysis results obtained that the factors that influence the prevalence of stunting in all districts/cities in NTT Province are the percentage of toddlers weighed ≥ 4 times, the percentage of toddlers receiving complete basic immunization, the percentage of households consuming iodized salt, the percentage with viable drinking water sources and the amount of expenditure per capita. Based on the results of the analysis, it is known that LCR-GWR with the Adaptive Gaussian Kernel weighted function is able to produce a better model than the GWR model in overcoming local multicollinearity problems in stunting in NTT Province, with the RMSE value (0.0344) smaller than the GWR RMSE model (3.8899).
- NTT Province based on the results of the 2013 Riskesdas and the 2016 and 2017 PSG was recorded as the province with the highest prevalence of stunting nationally. Efforts should be made to formulate policies that are integrated with spatial aspects in order to reduce the prevalence of stunting. The GWR model approach might be the right choice because it can overcome the problem of spatial heterogeneity. However, GWR is not able to accommodate multicollinearity problems that have the potential to occur at the local regression coefficient between predictor variables. Therefore the LCR-GWR model approach is used which is the development of the GWRR model, using locally compensated ridge, n ridge bias coefficients for N observation region adjusts to the effect of collinearity between predictor variables in each region. The use of LCR-GWR is expected to be able to accommodate the problem of local multicollinearity so that it can be obtained more accurate estimation of model coefficient parameters. From the analysis results obtained that the factors that influence the prevalence of stunting in all districts/cities in NTT Province are the percentage of toddlers weighed ≥ 4 times, the percentage of toddlers receiving complete basic immunization, the percentage of households consuming iodized salt, the percentage with viable drinking water sources and the amount of expenditure per capita. Based on the results of the analysis, it is known that LCR-GWR with the Adaptive Gaussian Kernel weighted function is able to produce a better model than the GWR model in overcoming local multicollinearity problems in stunting in NTT Province, with the RMSE value (0.0344) smaller than the GWR RMSE model (3.8899).