Confidence intervals and precise <I>p</I>-value in a scientific paper in a biomedical journal

Abstract

The International Committee of Medical Journals Editors recommends that the data obtained be quantified and presented with appropriate indicators of measurement errors. There is a widespread belief among professional statisticians that it is necessary to present precise p-values in publications without correlating them with any threshold (critical) values. The coincidence intervals are extremely informative. For continuous random variables, not only for the sample mean, but also for other indicators: median, standard deviation, asymmetry, kurtosis and coefficient of variation. For discrete random variables, not only for frequency, but also for most epidemiological indicators – the odds ratio, positive and negative predictive values. Regression analysis should end with the construction of confidence interval for all coefficients, as well as the indication of predictive intervals on the graph. In the case of assessing the “strength of the influence of the factor” in the analysis of variance, a confidence interval for this indicator is also obligate. Specifying the exact p-value obtained during the study increases the probability of the published article being included in the meta-analysis. For most calculations of the above-mentioned indicators, the MS Office package, plug-ins to it or calculators located on freely available Internet resources are sufficient.