A study about the size and power of t-Student and Wilcoxon tests - DOI: 10.4025/actascitechnol.v27i1.1495

• Autor(a) principal: Barros, Emílio Augusto Coelho
• Data de Publicação: 2008
• Outros Autores: Mazucheli, Josmar
• Tipo de documento: Artigo
• Idioma: por
• Título da fonte: Acta scientiarum. Technology (Online)
• Texto Completo: http://www.periodicos.uem.br/ojs/index.php/ActaSciTechnol/article/view/1495
• topic: tamanho do teste; poder do teste; simulação Monte-Carlo; teste t-Student; teste de Wilcxon; 3.00.00.00-9 Engenharias

A typical problem in statistics data analysis consists in testing a null hypothesis H0 : µ= µ0 versus an alternative H1 : µ≠ µ0 , where µ0 is some specific value of the true parameter µ. From a random sample with normal distribution with mean µ and variance σ2, both unknown, we might use the well known t-student test. As an alternative to the usual t-student test, under the symmetric supposition, we can use the nonparametric Wilcoxon on test (Conover, 1971). In this paper, a Monte-Carlo simulation study was conducted to calculate the empirical size and power of the t-student and Wilcoxon tests. In this study, several instances were considered. For the size of both tests, we considered sample sizes equal to n = 10, 20, K , 90, 100 simulated from Normal, Laplace, Uniform, t-Student and Logistic distributions, under the null hypothesis with σ = 0. In the power study, from the same sample sizes and for all distributions, random samples were simulated from the alternative hypothesis considering µ0 = -1.0, -0.9, K , 0.9, 1.0