Simulated Critical Differences for Speech Reception Thresholds

Purpose

Critical differences state by how much 2 test results have to differ in order to be significantly different. Critical differences for discrimination scores have been available for several decades, but they do not exist for speech reception thresholds (SRTs). This study presents and discusses how critical differences for SRTs can be estimated by Monte Carlo simulations. As an application of this method, critical differences are proposed for a 5-word sentences test (a matrix test) using 2 widely implemented adaptive test procedures.

Method

For each procedure, simulations were performed for different parameters: the number of test sentences, the j factor, the distribution of the subjects' true SRTs, and the slope of the discrimination function. For 1 procedure and 1 parameter setting, simulation data are compared with results found by listening tests (experimental data).

Results

The critical differences were found to depend on the parameters tested, including interactive effects. The critical differences found by simulation agree with data found experimentally.

Conclusions

As the critical differences for SRTs rely on multiple parameters, they must be determined for each parameter setting individually. However, with knowledge of the test setup, rules of thumb can be derived.

http://ift.tt/2j2kw4g