On K-Group Classification Procedures Using Precedence Probabilities

• Activity Number: 262
• Type: Contributed
• Date/Time: Monday, August 10, 2015 : 2:00 PM to 3:50 PM
• Sponsor: Section on Nonparametric Statistics
• Author(s): Rajarshi Dey* and Paul Nelson
• Companies: University of South Alabama and Kansas State University
• Keywords: K group classification ; Precedence Probability ; U statistic

Abstract

Consider F_{i}; i=1,2,...,K to be K separate distribution functions representing K groups and X_{i} to be an observable independent sample vector generated from F_{i}. The problem is to assign new observations to any of the K groups. Define, P[X_{i1}< X_{i2}< ...< X_{iK}] to be a precedence probability for K groups if (i1,i2,...,iK) is a permutation of K integers. Using all K! precedence probabilities we provide an efficient method for classification of new observations into one of the K groups. This procedure could also be used in case we have a multivariate sample vector.