Objective Response Detection in the Frequency Domain: Comparison of Several q-Sample Tests

Abstract

The objective assessment of the hearing threshold requires a suitable statistical test for response detection. For checking the q spectral lines of an auditory-evoked potential spectrum in the frequency domain, a q-sample test seems to be more favourable compared to q applications of a one-sample test. The response detection performance of three q-sample tests (q-sample analogue of Watsons U² test, q-sample uniform scores test and a modified q-sample uniform scores test) has been checked by extensive Monte Carlo simulations. To compare the performance of the three test statistics, sensitivity was calculated and receiver-operating characteristics were constructed from the probability density functions estimated by the Monte Carlo simulations. In addition, a comparison on the basis of real near-threshold auditory brainstem response data was carried out. The modified q-sample uniform scores test appeared to be the most powerful test. Some aspects of the practical application of this test are discussed in this paper.