ROBUST LEAST MEDIAN OF SQUARE MODELLING USING SEEMINGLY UNRELATED REGRESSION WITH GENERALIZED LEAST SQUARE ON PANEL DATA FOR TUBERCULOSIS CASES

Abstract

Tuberculosis, primarily affecting the lungs and other organs, was the leading cause of death worldwide before the COVID-19 pandemic and continues to be a significant health concern. This research examined tuberculosis (TB) using a panel dataset. As a consequence, the datasets may contain outliers and contemporaneous correlations. A Robust Least Median of Square (LMS) model was developed in this research by combining Seemingly Unrelated Regression (SUR) with Generalized Least Square (GLS) on panel data to provide an analysis overview to overcome outliers and contemporaneous correlations. Based on secondary data obtained from the Central Bureau of Statistics of the Gorontalo Province and the Ministry of Health of the Gorontalo Province, this research examines TB cases between 2017 and 2021. The Chow test result suggests that CEM is the most appropriate model for analyzing panel data for TB cases in Gorontalo Province between 2017 and 2021. Due to the presence of outliers and influential observations in the data, robust LMS is employed. Furthermore, there is a problem of contemporaneous correlation in this research. Each regency or city can mitigate this problem by implementing robust LMS using SUR with GLS.