Robust Estimations of Cluster Analysis: Practical Application in Administrative and Financial Corruption Abstract

Abstract

Cluster analysis (clustering) is mainly concerned with dividing a number of data elements into clusters. The paper applies this method to create a gathering of symmetrical government agencies with the aim to classify them and understand how far they are close to each other in terms of administrative and financial corruption by means of five variables representing the prevalent administrative and financial corruption in the state institutions. Cluster analysis has been applied to each of these variables to understand the extent to which these agencies are close to other in each of the cases related to the administrative and financial corruption. Outliers and infected data of the well-thought phenomenon have led to inaccurate results that were highlighted by the cluster analysis process made on the infected data. This gave rise to adopt efficient estimation methods known as the Robust Methods, which are used when the deliberate phenomenon-related data is infected due to certain outliers. Thus, this paper is purposed to obtain robust estimations that are functional in determining the robust distances for elimination of outlier and data cleansing by means of certain robust methods the stahel-donoho estimator.