Comparative Performance of Spatial Robust Small Area Estimation Methods: A Simulation Study

Arianingsih, Baiq Dian (Study Program of Statistics and Data Science, IPB University, Bogor, Indonesia)
Sadik, Kusman (Study Program of Statistics and Data Science, IPB University, Bogor, Indonesia)
Indahwati, Indahwati (Study Program of Statistics and Data Science, IPB University, Bogor, Indonesia)

Zero : Jurnal Sains, Matematika, dan Terapan
Vol 9, No 3 (2025): Zero: Jurnal Sains Matematika dan Terapan

Estimating parameters for small areas often faces limitations due to insufficient sample sizes, resulting in low-precision estimates. The Small Area Estimation (SAE) approach is used to address this problem by utilizing auxiliary variables to improve estimation efficiency. This study evaluates four SAE methods, namely EBLUP, REBLUP, SEBLUP, and SREBLUP, through a simulation study based on a nested error model across 18 scenarios that combine two area sizes (16 and 64 areas), levels of outlier contamination in the error component, and degrees of spatial correlation in the area-level random effects. Each scenario is replicated 50 times. Model performance is evaluated using Relative Bias (RB) and Relative Root Mean Square Error (RRMSE). The results show that non-robust methods are sensitive to outliers, whereas robust methods produce more stable estimates. The SREBLUP method demonstrates the best performance under low to moderate spatial correlation. In addition, an ANOVA test is conducted to identify factors that significantly affect the response.