Abstract
The variability within a collection of 40 upland rice genotypes, comprising plant introductions, accessions, and breeding lines was studied during the growth cycle, under upland growing conditions. Statistical evaluation of data was done using factor, discriminant, and cluster analyses in order to determine the proportion of characters to the variability and identify the genotypic sources of these characters for hybridization purposes. Genotypes were quite variable in agronomic and morphological characters. The first three axes of factor analysis accounted for about 47% of the total variance, while 79% of the variability within the genotypes was explained by the first three discriminant canonical functions. The two multivariate techniques differ in the characters identified as most important in describing the genotypes. They, however, jointly identified yield sink capacity, grain yield, panicle number per plant, days to flowering, panicle length, and grain length and width as the characters contributing most to the divergence among the genotypes. The genotypes were sorted into 10 clusters whose membership cut across different agro-ecological regions. Differences in the value of characters between clusters could be exploited for hybridization and further selection for increase in grain yield.
