Use of the nonparametric bootstrap in evaluating genotypic, phenotypic and environmental correlations - DOI: 10.4025/actasciagron.v30i5.5966

Abstract

This study was conducted to propose the use of the empirical nonparametric bootstrap procedure in order to test the significance of the correlations. Eight different population sizes and 10 characteristics were simulated at different correlation levels in each population. The efficiency of the bootstrap method was evaluated by comparing the results of the method, relative to a t-test. The bootstrap method provided identical results to those obtained by the t-test at 1% probability, for the population sizes of 25; 50; 100; 250; 500; 2500 and 5000, therefore showing the adequacy of 5,000 replicas. In general, the effectiveness of the bootstrap was not compromised by the size of the sample under study. The method showed high reliability, in addition to being an appropriate and useful procedure to be adopted in testing the significance of both genotypic and environmental correlations for the genetic improvement of multiple characteristics. The study of the correlation magnitudes that provided both type I and type II errors in all populations reveal the bootstrap method to be appropriate not only for testing the genetic and environmental significances of the correlations, but also in testing phenotypic correlations.