Modeling the Test-Retest Statistics of a Localization Experiment in the Full Horizontal Plane

Area of Science:

• Auditory Neuroscience
• Biomedical Engineering
• Speech and Hearing Science

Background:

• Sound localization in the horizontal plane is challenging for individuals with hearing impairments.
• Localization experiments are crucial for assessing cochlear implant (CI) recipient progress.
• Test-retest reproducibility in these experiments is limited by listening and time constraints.

Purpose of the Study:

• To introduce and evaluate two statistical models for analyzing test-retest data from auditory localization experiments.
• To assess the efficiency of Gaussian distribution and surrogate data models in describing directional hearing ability.
• To provide methods for comparing test-retest measurements in localization studies.

Main Methods:

• A localization experiment was conducted using a 12-loudspeaker setup in an anechoic chamber.
• Stimuli (HSM sentences) were presented pseudo-randomly from 12 directions at 65 dB.
• Data from 33 CI patients (92 test-retest pairs) were retrospectively analyzed using Gaussian distribution and surrogate data models.

Main Results:

• The first model, not accounting for direction inversions (e.g., front-back confusion), showed limitations.
• The second model incorporated inversions but was not universally applicable across all outcome measures.
• Both models successfully described the test-retest statistics of directional hearing.

Conclusions:

• The developed models effectively characterize test-retest variability in auditory localization.
• Performance differences between models across various measures preclude a general recommendation.
• The presented statistical framework facilitates paired test comparisons for localization experiments.