Proposal for evaluation method of proficiency test results on EMI measurement

Z-score based evaluation of proficiency testing for general tests is standardized in ISO 13528 [1]. Therefore, proficiency testing for EMI measurements is also evaluated with Z-score. Since Z-score is supposed to be calculated on the assumption that the distribution is normal distribution in the linear value, accordingly, the decibel values reported as EMI measurement readings shall be converted once to linear values before the calculation of Z-score. It has been revealed, however, that there are cases in which Z-score calculated in this process is undervalued if readings vary greatly. Therefore, we did verification by way of a simulated calculation about the Z-score to be obtained in calculation of raw decibel readings collected from participants. Data group we assumed for the simulation comes with certain dispersion in the linear value. The result of this simulated calculation was as follows. That is, the Z-score becomes inflated, for collected readings smaller than the median values, than the value calculated with linear value. This is because the distribution of decibel values is non-centric t-distribution. Their distribution, however, becomes closer to the distribution of measurement results, which implies that this method is appropriate for the evaluation of EMI testing.