Uso inercial de técnicas estadístico-inferenciales: su posible impacto en el hallazgo de resultados falsos en salud

Abstract

Objective: To evaluate the degree in which important statistical resources, particularly p-values, confidence intervals and procedures for determining sample sizes, are used in the biomedical literature through a ritual way. Methodology: A total of 25 original articles published in each of 4 journals indexed in the biomedical field were selected. For each of them, it was assessed whether they follow the instructions of the STROBE and CONSORT guidelines regarding the sample size, as well as the use of p- values, confidence intervals, and the use of these in the discussion of the article. Results: The 97.0% of the articles reported the sample size, but only 62.9% explained how it was determined. The p-value is used more frequently (68.0%) than the confidence intervals (63.9%). Only 15.5% uses confidence intervals in the discussion section. Conclusion: the most conventional statistical tolos are used more in ceremonial way rather than in a functional one.