The influence of threshold values on Type A standard uncertainty in mass flow rate measurements for liquids

In metrological studies, the minimization of Type A standard measurement uncertainty, which can be significantly increased by the presence of outliers in the data, is an essential task. Our previous research focused on detecting and eliminating outliers using various statistical methods, such as the Interquartile Range (IQR) method, the Median Absolute Deviation (MAD) method, as well as the machine learning-based Isolation Forest method. These methods have proven effective under certain conditions, significantly reducing Type A uncertainty and improving the measurement accuracy. However, despite the progress made, the issue of determining the optimal threshold values at which measurement uncertainty may well be minimized remains unresolved. This is especially crucial for tasks where even a slight increase in the uncertainty may significantly affect the results. The main focus is on studying the relationship between threshold values and changes in the measurement uncertainty, as well as analysing possible causes of these changes. The study is aimed at identifying the optimal data processing conditions under which minimal measurement uncertainty can be achieved, which is of paramount importance for improving the accuracy of metrological studies.