Comparación de potencias en pruebas estadísticas de normalidad, con datos escasos

Resumo

Verification that the observed data on a random phenomenon may be assumed to have been drawn from a Normal Probability Distribution is necessary in a variety of practical situations. Frequently in these problems, there are few observations. In the field of statistics many hypothesis tests are proposed to verify normality. Some researchers have conducted studies on most of the tests in question, including the determination of power. Unfortunately, verifications generally operate with samples of fifty or more data. Therefore, in this work the power of various tests is estimated with samples of ten, fifteen and twenty data. Besides, power is determined in situations where the original distribution is symmetrical, which undoubtedly represents the worst condition for the test. For these operations simulation experiments are performed. The paper concludes with a qualitative assessment of the appropriateness of the analyzed tests.