Confidence interval

Abstract

Confidence interval presents a range of possible values within which, with some certainty, we can find the statistical measure of the population. As such, it is an objective estimate of (in)precision and sample size of certain research. Therefore, we can consider confidence interval also as a measure of the sample and research quality. Confidence interval is defined by its margins of error. Depending on the confidence level that we choose, the interval margins of error and respective range also change. The most used confidence intervals in the biomedical literature are the 90%, 95%, 99% and not so often 99.9% one. The narrower the margins of an interval are, the higher is the estimate accuracy. The 95% confidence interval is traditionally the most used interval in the literature and this relates to the generally accepted level of statistical significance P < 0.05. There is a rule for same sized samples: the smaller the confidence level is, the higher is the estimate accuracy. Only the studies with a large sample will give a very small confidence interval, which points to high estimate accuracy with a high confidence level. A confidence interval can be attributed to almost every statistical measure. Although there are some other ways of calculating it, the confidence interval is generally and most frequently calculated using standard error. The P value and the confidence interval are two complementary statistical indicators. They describe the same thing, but in two different ways. The P value describes probability that the observed phenomenon (difference) occurred by chance, whereas the confidence interval provides margins of error within which it is possible to expect the value of that phenomenon. In the last twenty years, increasing number of journals require reporting of the confidence intervals for each of their key results. Reporting of this confidence interval provides additional information about the sample and the results. It is, moreover, very useful and irreplaceable supplement to a classical hypothesis testing and to the generally accepted P value. It should become a standard of all scientific journals to report key results with respective confidence intervals because it enables better understanding to the interested reader.