Methods of longitudinal data analysis for the genetic improvement of sugar apple

Abstract

The objective of this work was to compare the ways of analyzing repeated measures to improve the production of sugar apple (Annona squamosa). Twenty half‑sib progenies were evaluated, over three years (2003, 2004 and 2005), in a randomized block design with five replicates, and each plot was constituted of four plants. The evaluated trait was the number of fruit per individual. The models of compound symmetry, autoregressive with heterogeneous variance, the structured ante‑dependence, and compound symmetry with heterogeneous variance were analyzed using the ASReml software. The estimation of variance components and the prediction of breeding values were made by the REML/BLUP. The comparison of the models was done by the likelihood ratio test and Akaike’s information criterion. The structured ante‑dependence model, for the factors progeny and parcel, and the multivariate model, for the residual factor, are the best approaches for data analysis, providing efficiency and parsimony over the full multivariate model. With the structured ante‑dependence model, it is possible to identify superior families in each harvest, and also the families with larger total number of fruit.