Evaluation of frequency-weighted sound level measurements for prediction of low-frequency noise annoyance

The objective was to evaluate different frequency weightings for prediction of rated annoyance. In Study 1, cluster analysis was used to select two groups, which were exposed to low and high frequency noise, respectively. This was done to maximise the possibility of differentiating between frequency weightings. Correlations were calculated between rated annoyance and the different frequency-weighted values. Correlations did not indicate that dB(A) was inferior to any other frequency-weighted sound level as a predictor of annoyance. In Study 2, multiple and logistic regression analyses, using L-Aeq and the dB(C)-dB(A) difference as independent variables, and annoyance as dependent variable, were carried out in a group of 709 persons. The proportion of explained annoyance variance was increased by 1.4% when the dB(C)-dB(A) difference was added to the analysis. Also a further 2.4% of the annoyed persons were identified by adding the difference in the logistic regression. The dB(C)-dB(A) difference thus may make a significant, although small, contribution to the explanation of differences in annoyance ratings. (C) 1997 Elsevier Science Ltd.