Abstract

COSTA, Sérgio Francisco. Nível de significância alpha ou valor-p?. Est. Aval. Educ. [online]. 2002, n.25, pp.185-198. ISSN 0103-6831.

During the last few decades, some researchers have shown strong preference, when testing statistical hypoteses, for the effective level of significance, also called p-value, in place of the traditional theoretical level of significance alpha, established in advance - at the time of the design of the experiment. This article tries to show that such practice adds extreme flexibility to the critical region, leaving the question of rejecting or not rejecting the hypoteses under the almost inevitable influence of emotional factors. Some aspects in relation with the occurrence of Type II Errors as Well as the corresponding consequences connected with the sensitivity of the experiment (power) are dealt with.

Keywords : Level of Significance; Valor-p; Critical Region; Type I Error; Type II Error; Power.

        · abstract in Portuguese     · text in Portuguese     · Portuguese ( pdf )