The p-value is a widely used tool for inference in clinical studies. However, despite the numerous books and papers published on the basics of statistical inference and, thus, on the p-value, there still seems to be a need to highlight what message the p-value exactly contains (and what it does not). In this article, the basic concepts and the different misconceptions regarding the p-value will be highlighted and illustrated with a clinical trial in osteoarthritis. It will also be shown that the (95%) confidence interval is to be preferred over the p-value as a statistical inference tool.
Biostatistical Centre, Catholic University of Leuven, Leuven, Belgium. e.lesaffre@erasmusmc.nl
Bull NYU Hosp Jt Dis. 2008 Jan 1; 66 (2): 146-9.
AbstractThe p-value is a widely used tool for inference in clinical studies. However, despite the numerous books and papers published on the basics of statistical inference and, thus, on the p-value, there still seems to be a need to highlight what message the p-value exactly contains (and what it does not). In this article, the basic concepts and the different misconceptions regarding the p-value will be highlighted and illustrated with a clinical trial in osteoarthritis. It will also be shown that the (95%) confidence interval is to be preferred over the p-value as a statistical inference tool.