Classification of Stateless Communities Using a Robust Nonparametric Kernel Discriminant Function

Abstract

The main objective of this study was to classify the stateless communities using a Robust Nonparametric Kernel Discriminant Function. A Robust Nonparamet- ric Kernel Discriminant Function has therefore been developed by modifying the traditional using Bayes Discriminant Rule with a Nonparametric Kernel Discrim- inant Function. A suitable Kernel method was carefully chosen and a series of bandwidths were tested to get what could work best for our model. The study also estimated the Classification Rates of the developed function as a measure of its Robustness. The function was compared with parametric functions such as Linear Discriminant Function and Quadratic Discriminant Function through a simulation study. The result has been applied in classifying the stateless com- munities. As of today, the Pemba people in Kenya are among a number of other communities in the world which have been identified and listed as Stateless. Ac- cordingly, as a way of demonstrating how the Function works, it has been used to identify the Pemba who live in Kenya as stateless people, and then suggest integration of them into the Neighboring Giriama or Rabai Community based on displayed intersecting characteristics. In operationalizing the Robust Nonpara- metric Kernel Discriminant Function, data from the Kenya National Bureau of Statistics (KNBS) obtained from the 2009 Kenya Population and Housing Cen- sus and a survey report on Pemba Community conducted in 2015 was applied to the study. Various characteristics associated with the listed Tribes/Ethnic Com- munities such as Education Level, Religion, Housing Building Materials (Hous- ing Materials for the Floor, Walls and Roof), Waste Disposal, Source of Water and Employment Status, were considered. From the Theoretical developments and Empirical demonstrations, the findings from this study indicate that, the developed Nonparametric Discriminant Function provides a good classification method for classifying Stateless Communities. This is because they exhibit lower Misclassification Rates compared to the existing Parametric Methods. Use of the Kernel Discriminant Function is therefore recommended in classifying Stateless Persons. The study further recommends to the Government of Kenya to inte- grate the Pemba into either Giriama or Rabai communities and recognize them as Kenyan Citizens. Being that the methods developed and used herein are some- what Global, results from this study respond to a major push by United Nations Human Commissioner for Refugees to "map" the size of Stateless Populations and their Demographic Profiles and respective causes, potential solutions and associated Human Rights Situations. By classifying/associating Stateless Com- munities to a particular Local, yet already existing and properly defined/known xiii Community that is recognized, a way of integrating them is one of the poten- tial solutions, which then feeds into the greater Global Agenda regarding ending Statelessness across the world. This will help in making service delivery to such people without discrimination and go a long way in restoring their dignity.