Planejamento experimental e estratégias de análise para avaliação de progênies no melhoramento do feijoeiro
Loading...
Date
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Universidade Federal de Viçosa
Abstract
O feijão (Phaseolus vulgaris L.) se destaca entre as espécies de leguminosas, conhecidas popularmente como feijão, sendo o Brasil o maior produtor e consumidor. Diante da importância da cultura, os trabalhos de melhoramento de feijão têm visado o desenvolvimento de cultivares mais produtivas, com arquitetura de planta mais ereta e bom aspecto comercial de grãos. Nas fases iniciais dos programas de melhoramento, é comum a avaliação de um grande número de progênies. Em algumas espécies, como o feijoeiro por exemplo, normalmente ocorre restrição de quantidade de sementes para montar os experimentos, dificultando o uso de delineamentos com repetições. Uma solução é a utilização do delineamento de blocos aumentados (DBA). Entretanto, o DBA possui algumas desvantagens como, estimativa do erro experimental com base em cultivares testemunhas e ocupação de grande número de unidades experimentais com testemunhas. Uma alternativa seria a utilização do delineamento parcialmente replicado (P – REP), no qual uma porcentagem dos próprios tratamentos em avaliação é repetida, a fim de se estimar o erro experimental. Ainda são escassos na literatura, estudos que permitam uma comparação entre DBA e P – REP. Assim, no primeiro capítulo, avaliou-se a eficiência do P – REP em relação ao DBA na avaliação de linhagens de feijão. Já nos próximos experimentos, quando se tem maior disponibilidade de sementes, é comum a utilização de delineamentos mais robustos como o látice. Contudo, o uso de diferentes delineamentos nos ensaios sequenciais, ocasiona desbalanceamento estatístico dificultando a análise conjunta destes experimentos. Desta forma, propôs-se no segundo capítulo uma estratégia para conectar ensaios sequenciais em desbalanceamento estatístico. Uma outra opção que surge neste contexto de grande número de tratamentos em avaliação, escassez de sementes e desbalanceamento estatístico é a utilização do delineamento parcialmente replicado aumentado (A – PREP). No A – PREP uma porcentagem dos tratamentos é repetida nos diferentes ambientes de avaliação, de forma que ao se considerar todos os ambientes, todos os tratamentos apresentam o mesmo número de repetições, o que permite uma melhor estimativa dos efeitos de tratamentos. Isto posto, no terceiro capítulo avaliou-se a utilização do A – PREP em ensaios iniciais de avaliação de linhagens de feijão. Concluiu-se que: o P – REP foi mais eficiente que o DBA na avaliação da produtividade de grãos; o uso da modelagem de efeitos de blocos e resíduos admitindo a heterogeneidade de variâncias permitiu a conexão de ensaios sob desbalanceamento estatístico, promovendo uma seleção mais acurada; e que o uso do A – PREP é adequado para a avaliação inicial de linhagens de feijão, possibilitando melhor estimativa do efeito de tratamentos e, consequentemente, uma seleção mais acurada. Palavras-chave: Delineamentos de blocos incompletos; Acurácia seletiva; Desbalanceamento estatístico.
The common beans (Phaseolus vulgaris L.) stand out among legume species, commonly known as beans, with Brazil being the largest producer and consumer. Given the importance of this crop, bean breeding efforts have aimed at developing more productive cultivars with a straighter plant architecture and good commercial grain appearance. In the early stages of breeding programs, it is common to evaluate a large number of progenies. In some species, such as beans, there is often a restriction on the number of seeds available for experiments, making it difficult to use designs with replications. One solution is the use of augmented block design (ABD). However, ABD has some disadvantages, such as estimating experimental error based on control cultivars and occupying a large number of experimental units with controls. An alternative would be the use of partially replicated design (P-REP), where a percentage of the treatments being evaluated is repeated to estimate experimental error. Studies comparing ABD and P-REP are still scarce in the literature. Thus, in the first chapter, the efficiency of P-REP in relation to ABD was evaluated in the assessment of bean lines. In subsequent experiments, when there is a greater availability of seeds, it is common to use more robust designs. However, using different designs in sequential trials can lead to statistical imbalance, complicating the joint analysis. Therefore, the second chapter proposes a strategy to connect sequential trials with statistical imbalance. Another option that arises in this context of a large number of treatments under evaluation, seed scarcity, and statistical imbalance is the use of augmented partially replicated design (A-PREP). In A-PREP, a percentage of treatments is repeated across different evaluation environments, ensuring that when all environments are considered, all treatments have the same number of replications, allowing for better estimation of treatment effects. Thus, in the third chapter, the use of A-PREP was evaluated in the initial trials for assessing bean lines. The conclusions were: P-REP was more efficient than ABD in evaluating grain productivity; the use of block effect modeling and residuals accounting for variance heterogeneity allowed for the connection of trials under statistical imbalance, promoting more accurate selection; and the use of A-PREP is suitable for the initial evaluation of bean lines, enabling better estimation of treatment effects and, consequently, more accurate selection. Keywords: Incomplete block designs; Selective accuracy; Statistical imbalance.
The common beans (Phaseolus vulgaris L.) stand out among legume species, commonly known as beans, with Brazil being the largest producer and consumer. Given the importance of this crop, bean breeding efforts have aimed at developing more productive cultivars with a straighter plant architecture and good commercial grain appearance. In the early stages of breeding programs, it is common to evaluate a large number of progenies. In some species, such as beans, there is often a restriction on the number of seeds available for experiments, making it difficult to use designs with replications. One solution is the use of augmented block design (ABD). However, ABD has some disadvantages, such as estimating experimental error based on control cultivars and occupying a large number of experimental units with controls. An alternative would be the use of partially replicated design (P-REP), where a percentage of the treatments being evaluated is repeated to estimate experimental error. Studies comparing ABD and P-REP are still scarce in the literature. Thus, in the first chapter, the efficiency of P-REP in relation to ABD was evaluated in the assessment of bean lines. In subsequent experiments, when there is a greater availability of seeds, it is common to use more robust designs. However, using different designs in sequential trials can lead to statistical imbalance, complicating the joint analysis. Therefore, the second chapter proposes a strategy to connect sequential trials with statistical imbalance. Another option that arises in this context of a large number of treatments under evaluation, seed scarcity, and statistical imbalance is the use of augmented partially replicated design (A-PREP). In A-PREP, a percentage of treatments is repeated across different evaluation environments, ensuring that when all environments are considered, all treatments have the same number of replications, allowing for better estimation of treatment effects. Thus, in the third chapter, the use of A-PREP was evaluated in the initial trials for assessing bean lines. The conclusions were: P-REP was more efficient than ABD in evaluating grain productivity; the use of block effect modeling and residuals accounting for variance heterogeneity allowed for the connection of trials under statistical imbalance, promoting more accurate selection; and the use of A-PREP is suitable for the initial evaluation of bean lines, enabling better estimation of treatment effects and, consequently, more accurate selection. Keywords: Incomplete block designs; Selective accuracy; Statistical imbalance.
Description
Citation
PEREIRA JÚNIOR, José Domingos. Planejamento experimental e estratégias de análise para avaliação de progênies no melhoramento do feijoeiro. 2024. 105 f. Tese (Doutorado em Fitotecnia) - Universidade Federal de Viçosa, Viçosa. 2024.
