In a usual hierarchical cluster analysis, the identification of genetic distances among germplasm accessions through morphological data on qualitative as well as quantitative characters recorded using the International Plant Genetic Resources Institute (IPGRI)/ National Bureau of Plant Genetic Resources (NBPGR) descriptors, the high numerical nature of the quantitative data leads to the masking of smaller numerical but very important qualitative characters, leading to a lesser precision. This is true with D2 statistics and also the principal component based vector analysis leading to genetic divergence, restricting the use of highly significant qualitative parameters in germplasm characterization, often resulting in misclassifications even at the species level. Hence, the molecular markers are relied upon for final conclusions. Efficient data transformation systems to arrive at the exact genetic distances by accounting for both the qualitative as well as quantitative characters with equal weightage are being detailed hereunder. Among the models proposed, (value – mean)/SD was proved to be the best and the results are further supported by the factor analysis of principal components derived from the correlation matrix.

Augmented designs (Federer, 1956; and Sapra and Agarwal, 1992) are recommended for the evaluation of germplasm accessions through standardized morphological descriptors designed by the National Bureau of Plant Genetic Resources (NBPGR) (Srivastava et al., 2001), which act as tools for revealing the superiority of a particular accession over the existing ones with reference to the character of interest rather than revealing the cumulative genetic distances among the accessions. Cluster analysis (Falconer, 1989) as well as D2 statistics (Mahalanobis, 1928 and 1936) establish the comparative distance among the accessions by simultaneously considering the maximum possible morphological characters. Clustering based on a single character is proved to be exact (Sharma, 1978) in these techniques and the molecular support for the same was provided through randomly amplified polymorphic DNA analysis (Mathew, 2004). However, hierarchical clustering through morphological data on the maximum possible characters recorded using standard descriptors was proved to be less precise since the simultaneous usage of qualitative and quantitative characters leads to the masking of qualitative data by the latter with the weightage arising out of its higher numerical nature; for which D2 statistics is also not an exception (Chandrasekharaiah et al., 1969). The methodology for transformation of morphological data to avoid this problem and to ensure equal consideration for both in a hierarchical clustering is being discussed.

Thirteen pre-identified (stability of expression) accessions of French bean (Table 1) representing a wide genetic base were studied under cold arid desert conditions of Ladakh using the randomized block design with five replications. Following the NBPGR and International Plant Genetic Resources Institute (IPGRI) descriptors, the expressions on 51 morphological characters were recorded and the same were analyzed using the Statistical Package for Social Sciences (SPSS v.12). The results using the non-transformed data were studied in comparison to the morphological characteristics for formulation of transformation techniques, efficiency check and further confirmation through factor analysis of the principal components derived from a correlation matrix.

Genetic Distances,Germplasm Accessions, Falconer, National Bureau of Plant Genetic Resources, NBPGR, International Plant Genetic Resources Institute , IPGRI, Data Transformation Systems, Correlation Matrix, Statistical Package for Social Sciences , SPSS, Cluster analysis .