Two different indicators are commonly used to characterize accuracy of multiple measurements - standard deviation and confidence interval. This paper analyzes both indicators and shows that the confidence interval better expresses measurement accuracy. Usage of the standard deviation was unavoidable in the past, when there was no known technique to construct confidence intervals for indirect measurements with dependent arguments. Now this technique exists and has been used for almost half a century. Therefore, we conclude that presently, the confidence interval should be used as the only indicator of measurement accuracy in all cases of multiple measurements.
multiple measurements, standard deviation, confidence interval