A Statistical Method for the Analysis of Speech Intelligibility Tests.

Area of Science:

• Audiology and Hearing Science
• Statistical Modeling in Healthcare
• Speech Perception Research

Background:

• Speech intelligibility tests are crucial for evaluating hearing devices in hearing-impaired individuals.
• The traditional Speech Reception Threshold (SRT) is defined as the signal-to-noise ratio (SNR) for 50% correct speech perception.
• Conventional SRT analysis often uses adaptive procedures and repeated measures ANOVA, which may limit detailed performance insights.

Purpose of the Study:

• To introduce and validate an alternative statistical approach for analyzing speech intelligibility test data.
• To demonstrate the utility of a zero-and-one inflated beta regression model for speech perception analysis.
• To provide a more informative and interpretable method for assessing hearing device performance under various conditions.

Main Methods:

• Proposed a zero-and-one inflated beta regression model where individual sentence scores are the observations.
• Defined a model parametrization enabling efficient maximum likelihood estimation.
• Applied the model retrospectively to data from cochlear implant recipients, comparing results with traditional methods.

Main Results:

• The proposed model yielded mean SRTs consistent with traditional approaches but offered richer insights.
• Fitted psychometric curves revealed differential performance across SNRs, highlighting differences in slope.
• Results expressed as percent correct score differences were found to be more interpretable than traditional analyses.

Conclusions:

• The zero-and-one inflated beta regression model provides a more informative analysis of speech intelligibility data.
• This novel method enhances understanding of hearing device performance by detailing psychometric function slopes.
• The approach offers greater interpretability, particularly beneficial for evaluating speech perception in challenging listening environments.