CORRELATION AND PATH-COEFFICIENT ANALYSES OF SEED-YIELD COMPONENTS INPEARL-MILLET X ELEPHANT GRASS HYBRIDS

Correlation and path-coefficient analyses have been successful tools i n developing selection criteria. Since increased seed yield is an impo rtant goal in our pearl millet x elephantgrass [Pennisetum glaucum (L. ) R.Br. x P. purpureum Schum.] hexaploid breeding program, we used cor relation and path-coefficient analyses on seed data. This study was co nducted to develop appropriate selection criteria by determining the d irect and indirect effects of seed-yield components on seed yield plan t(-1). Number of tillers plant(-1), panicles tiller(-1), seeds panicle (-1), 100-seed weight, and seed yield plant(-1), were estimated for in dividual plants in seven families. Phenotypic (r(p)) and genetic corre lations (r(g)) were calculated, and path analyses (phenotypic and gene tic) were carried out according to predetermined causal relationships. Phenotypic and genetic correlations differed in several cases due to large environmental variance and covariance. Phenotypically, all compo nents were positively and significantly associated with seed yield pla nt(-1). Genotypically, only seeds panicle(-1) and 100-seed weight were significantly correlated. These two components were also positively c orrelated (r(p)=0.55, r(g)=0.63), so simultaneous improvement for both components would be feasible. Panicles tiller(-1) and seeds panicle(- 1) were negatively correlated (r(g)=-0.97). In the path analyses, all direct effects of the components on seed yield plant(-1) were positive . Phenotypic indirect effects were not as important as genetic indirec t effects. The components seeds panicle(-1) and 100-seed weight influe nced seed yield plant the greatest, both directly and indirectly.