<![CDATA[Direct estimation of the percentage of poor persons with disabilities at the regency/city level often has low precision due to limited sample sizes. This study compares Small Area Estimation (SAE) using Empirical Best Linear Unbiased Prediction (EBLUP) with logit transformation and Hierarchical Bayes (HB) with a Beta distribution across 35 regencies/cities in Central Java in 2025. Direct estimation yielded an average RSE of 42.86%, with 22 areas classified as poor and one area without an estimate. The application of SAE improved precision and produced estimates for all areas. EBLUP with logit transformation yielded an average RSE of 20.02%, with 30 areas classified as good, while HB Beta yielded an average RSE of 18.99%, with 31 areas classified as good. However, the random effects and residuals of the EBLUP logit model did not satisfy the normality assumption. Overall, HB Beta demonstrated better performance in terms of estimation precision and model suitability for proportional data.]]>
Copyrights © 2026
