On the uncertainty of sound reduction index measurements from inter-laboratory tests

Test reports of building acoustics measurements cannot be considered complete without declaring the associated uncertainty. Two dominant methods exist for calculating the un-certainty, i.e. the empirical approach, standardized in ISO 5725, and the modelling ap-proach, standardized in ISO/IEC Guide 98-3. The state of the art is so that this uncertainty must be assessed using the standard deviation of reproducibility obtained from multiple Inter-Laboratory Tests (ILTs), whose average values are standardized in ISO 12999-1. However, here three standard deviations are given: the usual repeatability and reproduci-bility standard deviations and an additional new standard deviation R95, neither defined nor used before. In the paper, it is shown that R95 causes an over-coverage in the fre-quentist sense, while the natural choice is the reproducibility standard deviation, both for laboratory and field measurements. Moreover, it is remarked that for products for which a specific inter-laboratory test (test code) already exists, a smaller uncertainty can be used, and this may produce an unbalanced situation on the market.