Reliability and repeatability of ISO 3382-3 metrics based on repeated acoustic measurements in open-plan offices

Abstract

This paper investigates variability in the key ISO 3382-3:2012 metrics, based primarily on the repeatability and reliability of these metrics, using repeated measurements in open-plan offices. Two types of repeated measurements were performed in offices – Type1 (n = 36), where the same path over workstations was measured from opposite ends, and Type2 (n = 7), where two different measurement paths were measured. Analyses performed per metric used (i) the range of observed values, i.e., ΔType1, ΔType2; and (ii) the observed values on their actual scales. Results from category (i) analysis: ΔType1‾, and bootstrapped 95% confidence intervals were 1.2 m (0.9,1.5) for distraction distance (rD); 0.8 dB (0.6,1.0) for spatial decay rate of speech (D2,S); 1.2 dB (0.8,1.5) for A-weighted sound pressure level of speech at 4 m (Lp,A,S,4 m); and 1.2 dB (0.7,1.7) for the A-weighted background noise level (Lp,A,B). ΔType2‾ were between twice and thrice the respective values of ΔType1‾. Results from category (ii) analysis: the reliability, based on intra-measurement correlation coefficients for repeated measurements, was fairly high for all metrics except for Lp,A,S,4 m for Type2 repeats. The repeatability limit/coefficient (r), which is the absolute difference between metric values not expected to be exceeded in 95% of the repeatability conditions, was 2.5 m for rD; 1.7 dB for D2,S; 2.9 dB for Lp,A,S,4 m; and 3.0 dB for Lp,A,B, for Type1 repeats. The r values for Type2 repeats were substantially higher except for D2,S; Lp,A,B not applicable in the current context. Overall, most of the Type1 results seem reasonable considering repeats were conducted in complicated room acoustic environments, while Type2 repeats would benefit from larger sample sizes in future studies. Some recommendations are outlined for the ISO 3382-3 methodology vis-à-vis Type1 and Type2 repeats, including future research directions that go beyond increased sample sizes.