Modeling the proportion of measles cases using sparse least trimmed squares

Abstract

Measles is a highly contagious disease and a health problem in several countries, including Indonesia. In 2022, Indonesia will experience an extraordinary situation (KLB) of measles cases, with the number of measles cases reaching 3,341 across 223 districts/cities. This data shows an increase of 32 times compared to 2021. North Sumatra is one of the provinces included in the list of regions and outbreak status, with 127 measles cases recorded in 2022. This study aims to find the factors that influence the number of measles cases in North Sumatra: one dependent variable, 34 independent variables, and 33 observations made up the study's variables. The data model chosen contains information on the percentage of measles cases linked to health, economic, human resource, and environmental variables. In addition, this study employs high-dimensional (data with many explanatory factors) data and includes outliers. Data with a large number of explanatory factors and outliers can be handled with LTS sparse analysis. The 34 independent variables were successfully chosen and reduced to 14 using the LTS sparse model. In addition, based on the and RMSE values ​​for model evaluation, sparse LTS shows satisfactory results compared with classical LASSO, with and RMSE values ​​for sparse LTS being 93.75% and 0.2933, respectively. Then, the and RMSE values ​​for LASSO are -62.4% and 2.1734. The government can use these elements to guide lowering the number of measles cases in North Sumatra.

Keywords

• measles, outliers, lasso, robust regression, sparse regression