Chapter 6 146 Discussion This study evaluated the reliability, agreement, and construct validity of the signal detection test in a population of Dutch locomotive engineers. Experiment 1 The results of the test-retest experiment indicated sufficient reliability and agreement in most driving conditions. The ICCs were fair to good in all, but two driving conditions. Poor ICCs were found in the driving conditions of the ICM (both DSD and ATP signal) and Loc1800 noise (DSDonly).This might be explained by the relatively small sample size of experiment 1. In a smaller sample, the impact of deviating values can be higher. Due to the relatively small range in SNR50 that was measured in experiment 1, we conclude that the reliability of the signal detection test is only moderate when used in normally-hearing subjects. When taking into account that the range in SNR50 as measured in experiment 2 was much larger, we expect the reliability of the signal detection test to be higher when used in hearing-impaired locomotive engineers. This should be confirmed by future research. Based on Bujang & Baharum (2017), we recommend a sample size of at least 22 for detecting ICCvalues of 0.5 and higher. Since poor ICCs and substantially higher SEMs were found in the driving conditions of the ICM (both DSD and ATP signal) and Loc1800 noise (DSD only), we recommend not to use the test for research or clinical purposes in these three driving conditions. In the ambient noise of all but one train cabin, the ICCwas higher for the ATP signal than for theDSDsignal.Thismight be related to the fact the ICC is sample-dependent (De Vet et al., 2006). It is easier to distinguish individuals in a heterogeneous sample than in a sample that ismore similar with regard to the characteristic under study.The standard deviations of the ATP measurements were higher than those of the DSD measurements. A larger standard deviation expresses more heterogeneity, and thus a higher ICC can be expected. This study only assessed the reliability and agreement in normally-hearing subjects. Therefore, the measurement errors and ICCs in a sample of hearing-impaired subjects should be investigated by future research. Experiment 2 The construct validity of the signal detection test is supported by its moderate levels of correlations to most conventional hearing tests, specifically to pure-tone
RkJQdWJsaXNoZXIy MTk4NDMw