Using the confidence interval of the mean to detect systematic errors in one analytical run