Estimating hearing protector attenuation for the second percentile using parametric statistics

Abstract

One method for reducing hearing protector attenuation data for multiple frequencies is a single number method such as the Noise Reduction Rating (NRR). Averaged data from ten subjects, three replications per subject, are combined with a subtractive two standard deviation (σ) correction in order to predict the 98% protection rate (second percentile). This study examined the errors incurred by assuming that the distribution of attenuation data is sufficiently normal to equate a −2σ correction to the protection achieved by the second percentile. Attenuation for a foam earplug was measured on 92 subjects of varying experience, according to ASA STD 1-1975, with one replication per subject for both a “subject” and “experimenter” fit. The second percentile single number attenuation was calculated from the actual distribution, and estimated by using a −2σ correction, as in the NRR. For the subject fit, the NRR underpredicted the second percentile attenuation by 1.6 dB. For the experimenter fit, it overpredicted protection by 2.2 dB. The skewness and kurtosis of the distributions were also evaluated. Typical values for subject fit were g1 = 1.65, g2 = 0.37 al 500 Hz, and g1 = 1.18, g2 = 0.03 at 1 kHz. These findings indicate that for adequate sample sizes and subject selection criteria, parametric statistics may be used to describe the attenuation distributions.