Confidence intervals for the mean of sound exposure levels.

Abstract

This text transposes standard statistical estimators of the mean and its confidence intervals to the field of occupational noise exposure, assuming that the samples are independent and lognormally distributed. The hypothesis of lognormality is applied to the values of A-weighted sound exposure as defined in ISO 1999 and expressed in Pa(2).h, which is equivalent to the currently accepted hypothesis of normality applied to the values of equivalent continuous A-weighted sound pressure level L(Aeq,T) expressed in dB(A). By expressing the initial data and results in dB(A), the text provides an unbiased estimator of the mean sound exposure level and tables of confidence intervals according to the sample size n and the standard deviation S(L) of the L(Aeq,T) measured values. The values of the following confidence intervals are given: exact one-sided upper and lower 95 and 97.5% confidence intervals and exact two-sided symmetrical 95% confidence interval. These results are compared with the approximate two-sided symmetrical 95% confidence interval proposed in standard, ISO 9612. This comparison demonstrates that the use of the approximate confidence interval can markedly underestimate the upper limit of the confidence interval when n is small and if S(L) is above 3 dB(A).