Desempenhos dos fatoriais fracionados em estimar efeitos principais na presença de interações duplas

• Autor(a) principal: Damasceno, Luiz Carlos Medeiros
• Data de Publicação: 2011
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: LOCUS Repositório Institucional da UFV
• Texto Completo: http://locus.ufv.br/handle/123456789/4043
• topic: Fatoriais fracionados; Delineamento De Cotter; Um Fator Por Vez; Fractional factorial; The design of Cotter; Um Fator Por Vez; CNPQ::CIENCIAS EXATAS E DA TERRA::PROBABILIDADE E ESTATISTICA::ESTATISTICA::PLANEJAMENTO DE EXPERIMENTOS

This study evaluated, through data simulation, the performance of fractional factorial designs of resolutions III, IV and V, the design of Cotter (1979) and UFPV method to estimate the main effects of the factors studied on an answer, presence or absence of only double interactions between them. For this, we estimated the main effects of three, four, five and ten factors with two levels. In the study of three factors, the main effects were estimated by the factorial and the UFPV. In order to study the four factors, the main effects were estimated by the fractional factorial, the UFPV and the design of Cotter (1979). For five factors were estimated by the main effects fractional factorial and at UFPV and the design of Cotter (1979). For the study of the ten factors, the main effects were estimated by the fractional factorial and at UFPV and the design of Cotter (1979). In all cases, 100 simulations were performed. To evaluate the performance of proposed designs, the main effects were estimated and applied separately for each situation, the Student t test for an average of 5% probability with 100 repetitions, according to the following hypothesis: Ho: p = parameter defined for each main effect separately. After analyzing the results, it was concluded that under the double presence of interactions, the fractional factorial resolution III UFPV and the method should not be used. Since the fractional factorial designs of resolutions IV and V and the design of Cotter (1979) study were satisfactory for three, four, five and ten factors, with or without the presence of double interactions.