key: cord-0982355-4w645rj6
authors: Ramos-Vera, Cristian Antony
title: The Bayes Factor, a Suitable Complement beyond Values of p<0.05 in Nursing Research and Education
date: 2021-03-05
journal: Invest Educ Enferm
DOI: 10.17533/udea.iee.v39n1e14
sha: 554b4eaa57e3b0b9caa1f7052d9624156c25f011
doc_id: 982355
cord_uid: 4w645rj6

nan

In accordance with a recent study in the current journal, significant differences are reported on burnout related with COVID-19 in two groups of nurses with and without experience with patients with COVID-19 infection in Iran, by using Student's statistical t test, (1) It reports that experienced frontline nurses exposed to treating COVID-19 patients indicate higher levels of job stress and burnout. This comparative analysis is among the most used in medical sciences based on the null hypothesis significance test (NHST) according to the "p <0.05" significance level that infers rejection of the null hypothesis (no difference) and provides greater likelihood confidence to researchers to assume the alternate hypothesis (difference) given the study sample. ( Bayesian statistics also permits contrasting hypotheses through probabilities of credibility, being a suitable complement to reinforce statistical significance and, when having frequentist significant findings, it is also a methodological alternative of statistical replication. (3) (4) (5) From the Bayesian model, the Bayes factor is the inclusive method of a priori and posteriori credibility to evaluate beyond the level of significance, given that it estimates the degree to which the data support the statistical hypotheses, from Jeffreys' classification scheme for Student's t analysis (5, 6) "weak", "moderate", "strong", and "very strong" ( Table 1) .

When having prior significant results, the only requirement is the t value (-2.86), as well as the sample sizes of both groups (151 and 94) reported by Hoseinabadi et al. (1) Regarding the Bayes factor, it permits inferring two interpretations:

FB 10 (in favor of the alternative hypothesis of significant difference) and BF 01 (in favor of the null hypothesis from lack of significant difference) and the 95% confidence interval. (4) Upon the evidence of significant difference, this analysis focuses on estimating the degree of certainty of the alternate hypothesis. The results obtained through the Bayes factor are BF 10 = 6.529 and BF 01 = 0.153, with 95%CI [-0.388 to -0.068]. These findings report moderate evidence in favor of the statistical hypothesis of significant difference; this may be interpreted in that the alternate hypothesis is six times greater than the nullity of the data and it -in turn -is reduced proportionally for some possible interpretation. Likewise, the parameter is reported of the maximum Bayes factor (maxBF 10 = 8.444) to determine the stability of the results, which indicates a greater support magnitude to the statistical differences, endorsing the reliability of the findings obtained. ( which only has a pair of recent guides that permit disseminating the use and interpretation of the Bayes factor beyond nursing research, encompassing its relevance in the health sciences in general. (5) (6) (7) Furthermore, it permits reinforcing systematic quantitative research that use said statistical tests, thus, providing greater inferential property to meta-analytical studies, and becoming an important methodological contribution inclusively for future articles in this journal.

Burnout and its influencing factors between frontline nurses and nurses from other wards during the outbreak of Coronavirus Disease (CO-VID-19) in Iran

Evidence against vs. in favour of a null hypothesis

Best uses of p-values and complementary measures in medical research: Recent developments in the frequentist and Bayesian frameworks

Bayesian reanalyses from summary statistics: a guide for academic consumers

Bayesian alternatives to null hypothesis significance testing in biomedical research: a non-technical introduction to Bayesian inference with JASP

Bayesian and frequentist testing for differences between two groups with parametric and nonparametric two-sample tests

Bayesian alternatives for common null-hypothesis significance tests in psychiatry: a non-technical guide using JASP