ROBUSTNESS AND POWER OF THE ONE MEAN T – STUDENT TEST AGAINST PRESENCE OF OUTLIERS.
Previous studies reveal that samples with outliers alter the type I and type II error of a t-Student test for inference of a mean. The methodology that these works use to simulate extreme data consists of mixing two different normal in order to contaminate the data. We think that this technique is n...
| Autores: | , , |
|---|---|
| Tipo de recurso: | artículo |
| Estado: | Versión publicada |
| Fecha de publicación: | 2020 |
| País: | Ecuador |
| Institución: | Escuela Superior Politécnica de Chimborazo |
| Repositorio: | Revista Perfiles |
| Idioma: | español |
| OAI Identifier: | oai:ojs2.localhost:article/70 |
| Acceso en línea: | https://perfiles.espoch.edu.ec/index.php/perfiles/article/view/70 |
| Access Level: | acceso abierto |
| Palabra clave: | outliers t-Student mean inference Probability atípicos media inferencia Probabilidad |
| Sumario: | Previous studies reveal that samples with outliers alter the type I and type II error of a t-Student test for inference of a mean. The methodology that these works use to simulate extreme data consists of mixing two different normal in order to contaminate the data. We think that this technique is not the most appropriate, since, when making this process, the result is not a new normal, which is breaching the main assumption of the test. In this work, this methodology is repeated in order to verify the problems described, but atypical data are also generated from a single normal without the need for any contamination. Using this last methodology, and with a stochastic simulation process, the probability of type I and type II error is estimated, from which it is concluded that the t-Student is a robust test against the presence of outliers and its power does not depend of the number of extreme data generated in the sample. |
|---|