A comparative performance of clustering procedures for mixture of qualitative and quantitative data – an application to black gram

Abstract

Knowledge of the genetic diversity of germplasm of breeding material is invaluable in crop improvement programmes. Frequently, qualitative and quantitative data are used separately to assess genetic diversity of crop genotypes. While assessing diversity based on qualitative and quantitative traits separately, there may occur a problem when the degree of correspondence between the clusters formed does not agree with each other. This study compares five different procedures of clustering based on the criterion of weighted average of observed proportion of misclassification in black gram genotypes using qualitative, quantitative traits and mixture data. The INDOMIX- and PRINQUAL-based clustering procedures, i.e. INDOMIX and PRINQUAL methods in conjunction with the k-means clustering procedure, show better performance compared with other clustering procedures, followed by clustering based on either quantitative or qualitative data alone. The use of the INDOMIX- and PRINQUAL-based procedures can help breeders in capturing the variation present in both qualitative and quantitative trait data simultaneously and solving the problem of ambiguity over the degree of correspondence between clustering based on either qualitative or quantitative traits alone.