PREHĽADNÉ ČLÁNKY * REVIEW ARTICLES

Confidence and tolerance intervals — a tool for biomedical data analysis aimed at clear evidence MIROSLAV MIKULECKY Bratislava, Slovak republic MIKULECKY M. Confidence and tolerance intervals – a tool for biomedical data analysis aimed at clear evidence. Cardiol 2004;13(4):211–215 In domestic as well as foreign publications standard deviation or standard error is often used to express statistical uncertainty, sometimes even as an undefined number. These indices possess a probabilistic interpretation only indirectly, after respecting the number of measurements. That is why they have to be substituted e.g. by intervals of 95 % confidence for mean and by those of 95 % tolerance for the individual. As the main tool of evidence the p value is usually used. Its information value, however, represents a mutually indistinguishable mixture of the effect size and its precision. Besides, it leads to the false impression of a lower risk than there really is. The best characterization of the size of the effect and of its precision is the confidence (for mean) or tolerance (for individual) interval of the difference as compared with the null value on one side, and with the minimal acceptable difference on the other. Using the p value, the replication probability – i.e. saying that a repetition of an observation or experiment will yield the same result – can be calculated. This probability makes the interpretation of p value appropriate. Adhering to old, no longer suitable schemes of statistical description can be explained by the lack in the education in biometrics at pre-and postgraduate level of medical teaching. Orientation towards inferential, conjectural statistics and arriving at conclusions together with their probability interpretation is needed. This knowledge is recently considered to be one of literacy – the ability to “read ” the sense of numbers. A physician-researcher and physician in practice will, for example, more effectively communicate with a drug producer, to the advantage of all four participants of the process, including the patient