In breeding programs, it is common to the assessment of many traits in order to practice selection in several of them simultaneously, in order to obtain uniform genotypes for defferent traits under selection. A viable alternative to this technique is the use of selection indices, using multivariate techniques to associate information about various traits of agronomic interest with the genetic properties of the studied population. The objective of this study was to predict the expected progress with selection by means of different multivariate selection indices. Were planted in October 2007, 140 full-sib progenies of passion fruit in square lattice design with two replications and plots consisting of three plants. We evaluated eleven characters. The non-parametric multivariate indices Mulamba and Mock and index the distance Genotype-ideotype, allowed to predict higher earnings and a balanced way between the characteristics when compared with the parametric indices of Smith and Hazel and Pesek and Baker. Using non-parametric indices was possible to select superior progenies for most traits simultaneously continuing the regional improvement program running.
Passiflora edulis; multivariate index; variability; response to selection