Detection of Towns Having a Peculiarity by Using Regression Models

Abstract

This paper proposes a method to detect towns having a peculiarity, which is a statistical outlier from a statistical table. A statistic often contains data that are peculiar and are also known as outliers which are followed as large residuals in regression models. The detection of outliers in statistical tables was studied. The table has 22 explanatory variables, one response variable and 1947 records which can clarify their efficient causes or mixed effects. This information have greatly helped local governments with their policy and improvement of each region, for example; infrastructures, public services, and subsidies or grants. Although many studies have been made on grouping records or building a predictive model to overcome outliers, little attention has been given to find outliers. Many of those studies require a model’s parameter tuning and learning, or a description of a fitting function. Furthermore, for municipal officers to find outliers, it would be desirable to be able to analyze readily Free Software R without programming. Therefore, we propose a method to detect outlier from a statistical table by using three regression models which do not require learning and parameter adjustment provided by R.