Impacto dos contrastes de interesse no delineamento de um experimento em blocos

AUTOR(ES)

Nilda Doris Montes Villanueva

DATA DE PUBLICAÇÃO

1993

RESUMO

Pearce s conceptions on statistical planning of experiments are explored in this work. The construction process of block designs was especially considered, in order to obtain experimental designs that depend upon the experimenter s needs. This way, the work is oriented basically to the study of contrasts of chief interest, because they are directly related with the objectives of the experimento Firstly, are studied the natural and basic contrasts. They are estimated independently and with effective replications and/or efficiency factors that are readily determined. Both indicate possible partitions of the treatment sum of squares and aid of the appropriate analysis of variance. The efficiency of most non-orthogonal designs proposed by Pearce (1963) is explicitly calculated using the spectral decomposition of the associated information matrix. From these studies were contructed a new table for the compilation of the results obtained. Next, is explored Gupta s theorem, which gives the relationship between the set of contrasts estimated with specified variances and covarianices and the information matrix of the design. This result allows us to find experimental designs compatible with the experimenter s specifications for the precision of estimation of contrasts of interest. Consequences and limitations of these results are examined for different struct ures of the variance-covariance matrix associated with the contrasts, especially for experiments for the comparison of test treatments with a control. Hypothetical experimental cases based on Pearce s example (1963) are shown also. Finally, it is studied the A-optimality of type S designs for the estimation of supplemented contrasts, extending Gupta s results (1989) for non-binary designs. The usefulness of the theory presented in the construction of designs is illustrated in severaI examples, using the statistical packages SAS and SOC for calculation of the relevant quantities

ASSUNTO(S)

planejamento experimental decisão estatistica

Documentos Relacionados

• Simulação de modelos mistos no delineamento em blocos aumentado
• Testes para verificar a igualdade de parâmetros e a identidade de modelos de regressão não-linear em dados de experimento com delineamento em blocos casualizados
• Estimação de contrastes de médias de tratamentos, de um experimento em blocos ao acaso, utilizando as análises clássica e espacial
• ESTRATÉGIAS DE MODELAGEM NA ANÁLISE DE EXPERIMENTOS EM DELINEAMENTO DE BLOCOS AUMENTADOS EM TESTES CLONAIS DE Eucalyptus spp.
• Comparação de tipos de análises em experimento com cenoura: látice quadrado versus blocos casualizados