Métodos de análise de dados longitudinais no melhoramento genético da pinha (Annona squamosa L.)

• Autor(a) principal: Mariguele, Keny Henrique
• Data de Publicação: 2010
• Tipo de documento: Tese
• Idioma: por
• Título da fonte: LOCUS Repositório Institucional da UFV
• Texto Completo: http://locus.ufv.br/handle/123456789/1323
• topic: Modelo misto; Matrizes de variância e covariância; Valor genético; Mixed model; Variance and covariance matrices; Genetic value; CNPQ::CIENCIAS BIOLOGICAS::GENETICA::GENETICA QUANTITATIVA

The pine cone or custard apple (Annona squamosa L.) is an important fruit crop in Brazil, especially in the Northeast. In this species, characters are repeatedly evaluated during the life of the organism and are called infinitely dimensional, in the sense that in each unit of time or age, the character can be measured, generating a set of multidimensional data. The interest in the analysis of such data usually resides in predicting values of individuals and progeny for particular point in time or through all points and in the identification of a parsimonious variance structure over time. The alternative of modeling can be applied to various random factors of the statistical model base. In the context of statistical experimental treatments considered fixed effects, these models, generally apply only to waste. But in case of improvement, in which treatments are considered genetic random effects, such modeling can be applied to the residual effects and also genetic. Also, different modeling (repeatability model, compound symmetry - CS, autoregressive with heterogeneous variance - ARH, ante-dependence structured - and multivariate SAD) can be applied to genetic effects and residuals, for example, modeling of genetic effects as ARH or SAD and mistakes by a multivariate model. This paper compares alternative ways of analyzing repeated measures to improve production of sugar apple. Twenty half-sibling progenies were evaluated over three years (2003, 2004 and 2005) in a randomized block design with five replications Each plot consisted of four plants. The trait was the number of fruits per individual. All models were analyzed using the software ASREML with the estimation of variance components and prediction of breeding values made by the REML / BLUP. The comparison of the models was the likelihood ratio test and the Akaike information criterion. Choosing the best model based on different covariance structures proved to be necessary in the analysis of repeated measures to maximize the efficiency of genetic improvement of cone. The SAD model for the factors and multiple offspring and parcel to the waste factor proved to be the best approach for data analysis, providing efficiency and parsimony over the full multivariate model. With the DSS model has been possible to identify superior families for each harvest and also with larger total numbers of fruit.