The effect of the volume of tests and the number of participants on the assessment of qualification in interlaboratory comparison tests

Abstract

The results of a comparative analysis of statistical indicators for assessing the proficiency determined during interlaboratory comparison tests (ICT) are presented. The main statistical indicators were selected: the assigned value, standard deviation for assessment the proficiency and standard uncertainty of the assigned value. Statistical indicators were determined in accordance with Algorithm A, GOST R 50779.60. Comparison of the indicators was carried out on the basis of numerical experiment data (random number generation) for the ultimate tensile strength of a D16AT aluminum alloy depending on the number of samples and the number of participants in the ICT qualification program. The calculation and subsequent statistical analysis of the proficiency test scores was based on the values generated by the software. Random number generation is performed for a characteristic that has a normal distribution law with the parameters of the mean equal to 450 MPa and standard deviation equal to 5 MPa. The values of the normal distribution parameters chosen as an example corresponded to the average level of tensile strength values of standard specimens made of D16AT aluminum alloy sheets. The numerical experiment (generation of random values) was carried out for given cases of the number of participants and the number of proficiency testing samples (PTS) independently of each other using separate samples of the appropriate size. A total of 36,000 ultimate tensile strength values were generated for the proficiency testing samples made of D16AT aluminum alloy sheets. As a result of calculations, 48 average values were determined (according to the number of considered implementations of ICT programs, depending on the number of participants and the proficiency testing samples) for each selected statistical indicator of the ICT qualification program. The average value of each statistical indicator was determined on samples with a bulk of 250 to 2000 generated (experimental) values, depending on the number of participants and the number of proficiency testing samples. The analysis thus performed made it possible to evaluate the influence of the volume of tests and the number of participants on the assessment of the proficiency in interlaboratory comparative tests. It is shown that for the selected range of the number of ICT participants (from 5 to 20) and the number of PTS (from 5 to 10), the value of the assigned x pt value does not depend either on the number of tested PTS or the number of ICT participants. The maximum discrepancy between the assigned value and the given level of tensile strength (450 MPa) is 0.13%, which falls within the error (rounding) for this level of values and cannot lead to significant errors in proficiency testing during ICT. The standard deviation for the qualification assessment, regardless of the number of participants, decreases with an increase in the number of proficiency testing samples, but such a decrease is insignificant and does not exceed the standard deviation (SD) value of the ultimate tensile strength (for the selected model for generating experimental values, the SD value is accepted to be 5 MPa). In contrast to the assigned value and the standard deviation for the qualification assessment, it is shown that the standard uncertainty of the assigned value for the ultimate tensile strength of standard samples made of D16AT aluminum alloy sheets depends on the number of participants and the number of proficiency testing samples tested by each participant and decreases with an increase in the number of participants in ICT programs and the number of proficiency testing samples.