Open AccessTo test or to estimate? P ‐values versus effect sizes
Author(s) -
Dunkler Daniela,
Haller Maria,
Oberbauer Rainer,
Heinze Georg
Publication year - 2020
Publication title -
transplant international
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.998
H-Index - 82
eISSN - 1432-2277
pISSN - 0934-0874
DOI - 10.1111/tri.13535
Subject(s) - confidence interval , medicine , observational study , statistics , statistical significance , hazard ratio , statistical hypothesis testing , p value , relative risk , value (mathematics) , econometrics , mathematics
Summary Most research in transplant medicine includes statistical analysis of observed data. Too often authors solely rely on P ‐values derived by statistical tests to answer their research questions. A P ‐value smaller than 0.05 is typically used to declare “statistical significance” and hence, “proves” that, for example, an intervention has an effect on the outcome of interest. Especially in observational studies, such an approach is highly problematic and can lead to false conclusions. Instead, adequate estimates of the observed size of the effect, for example, expressed as the risk difference, the relative risk or the hazard ratio, should be reported. These effect size measures have to be accompanied with an estimate of their precision, like a 95% confidence interval. Such a duo of effect size measure and confidence interval can then be used to answer the important question of clinical relevance.