Abstract
AbstractBackgroundDespite the efforts of leading statistical authorities and experts worldwide and the inherent dangers of interpretative errors in clinical research, misuses of statistical significance remain a common practice in the field of public health. Currently, there is a need to attempt to quantify this phenomenon.Methods97 studies were randomly selected within the PubMed database. An evaluation scale for the interpretation and presentation of statistical results (SRPS) was adopted. The maximum achievable score was 4 points. The abstracts and the full texts of the manuscripts were evaluated separately to highlight any differences in presentation. In this regard, a paired Student t-test was implemented.ResultsAll studies failed to adopt statistical significance as a continuous measure of compatibility between the model and the test result. The vast majority of them did not provide information on the validation of the model used. However, in most cases, all results were reported in full within the manuscripts. Substantial differences (a-f) between abstracts (a) and full texts (f) were highlighted when applying the SRPS (null hypothesis, P<.0001, a-f=1.2, Cohen’s D=1.7; a-f=0.5 hypothesis, P<.0001, a-f=0.7, Cohen’s D=1.0).ConclusionWhen contextualized in the current scenario, these findings provide evidence of widespread and severe shortcomings in the use and interpretation of statistical significance measures in clinical and public health research during 2023. Therefore, it is essential for academic journals to compulsorily demand higher scientific quality standards.
Publisher
Cold Spring Harbor Laboratory