Two approaches to cluster variable quantification

Subtitle:

Article type: Research Article

Authors: Dronov, Sergei V.^* | Sazonova, Antonina S.

Affiliations: Faculty of Mathematics and Information Technologies, Altai State University, Barnaul, Russia

Correspondence: [*] Corresponding author: Sergei V. Dronov, Faculty of Mathematics and Information Technologies, Altai State University, 61, Lenina, Barnaul 656049, Russia. Tel.: +7 906 963 9372; Fax: +7 3852 36 7067; E-mail: dsv@math.asu.ru

Abstract: Suppose our data is separated into some clusters. The problem of assigning the proper numerical labels to the clusters is considered. Referring to some medical applications, we suggest two methods to do it: first, when the labels are properly aligned with the numerical characteristics forming the clusters and second, oriented to the best adequacy of the linear regression model. Some attempts were made to compare these methods with each other and with the existing methods to solve similar problems.

Keywords: Quantification, categorized data, categorical linear regression, clusterization

DOI: 10.3233/MAS-140314

Journal: Model Assisted Statistics and Applications, vol. 10, no. 2, pp. 155-162, 2015