p-Hacking by Post Hoc Selection With Multiple Opportunities: Detectability by Skewness Test?: Comment on Simonsohn, Nelson, and Simmons (2014).

Simonsohn, Nelson, and Simmons (2014) have suggested a novel test to detect p-hacking in research, that is, when researchers report excessive rates of "significant effects" that are truly false positives. Although this test is very useful for identifying true effects in some cases, it fails to ident...

Descripción completa

Detalles Bibliográficos
Publicado en:Journal of Experimental Psychology. General Vol. 144; no. 6; pp. 1137 - 1146
Autores principales: Ulrich, Rolf, Miller, Jeff
Formato: Artículo
Publicado: American Psychological Association Dec2015
Materias:
Acceso en línea:Ver este registro en EBSCOhost
Descripción
Sumario:Simonsohn, Nelson, and Simmons (2014) have suggested a novel test to detect p-hacking in research, that is, when researchers report excessive rates of "significant effects" that are truly false positives. Although this test is very useful for identifying true effects in some cases, it fails to identify false positives in several situations when researchers conduct multiple statistical tests (e.g., reporting the most significant result). In these cases, p-curves are right-skewed, thereby mimicking the existence of real effects even if no effect is actually present.