An evaluation of statistical errors in clinical research articles published in physical medicine and rehabilitation journals

This study aims to systematically analyze the types of statistical errors encountered in clinical research articles published in the field of physical medicine and rehabilitation (PM&R) and to compare their distribution according to journal indexing types (SCI-E and TR Index). A total of 433 articles published between January 2024 and January 2025 were included in the evaluation; 82.7% of these were indexed in SCI-E and 17.3% in TR Index. The study evaluated multiple criteria, including the appropriateness of descriptive statistics, identification and proper selection of statistical tests, assessment of normality assumptions, application of post-hoc analyses when necessary, implementation of multiple comparison corrections, reporting format of p-values, inclusion of effect size and power analysis, as well as the reporting of statistical software used and the presence of ethics approval and data privacy statements. According to the analysis results, 49.4% of the reviewed articles reported p-values in closed form (e.g., p <0.05), and in 19.9% of the studies, parametric or non-parametric tests were applied without conducting a normality test. Multiple comparison corrections were omitted in 38.6% of the articles, and power analysis was not reported in 47.1% of the cases. Effect size was reported in only 51% of the articles. These findings indicate that structural deficiencies in statistical planning, implementation, and reporting remain widespread in clinical research publications. When comparing indexing types, TR index articles demonstrated significantly higher rates of normality testing (TR: 91.7%; SCI-E: 75.7%; p = 0.003) and power analysis reporting (TR: 68.6%; SCI-E: 47.7%; p = 0.001). In contrast, effect size reporting was significantly more common in SCI-E indexed articles (54.2%) than in TR Index (36.0%) (p = 0.004). Furthermore, SCI-E publications were more likely to include statistical software and version details and to report p-values in open form. These findings suggest that TR Index articles tend to prioritize pre-analysis planning steps, while SCI-E publications exhibit higher standards of statistical reporting quality and methodological transparency. The overall findings of the study reveal that statistical analyses in PM&R publications require substantial improvement, not only in terms of technical application but also in terms of adherence to reporting standards. In this context, widespread and mandatory use of internationally recognized reporting guidelines such as SAMPL, CONSORT, and STROBE is essential to enhance methodological integrity. Moreover, ethics committees should move beyond merely granting approval and take on a more active role in reviewing the methodological rigor of studies—a conclusion supported by the present findings. This thesis aims to serve as a guiding resource for researchers, journal editors, and peer reviewers in the PM&R field to promote higher standards of statistical quality in scholarly publications