Alyssa Everett

Alyssa Everett, AuD, CCC-A

University of Arizona
Department of Speech, Language, and Hearing Sciences

Predicting hearing aid success and determining which microphone mode to use prior to the fitting is an area of research that needs further investigation.

The current study examines possible predictive factors of hearing aid directional benefit. Specifically, we aim to answer the following questions,

1) Does an unaided speech-in-noise test or unaided acceptance of noise test predict a significant amount of variance in hearing aid directional benefit?

2) Do self-reported measures predict a significant amount of variance in the results on a speech-in-noise test, an acceptance of noise test, or directional benefit?


  1. Need to determine the mean differences between collocated and spatially separated conditions (4 different signal-to-noise ratios) for each measure (omni-directional and directional) [within each measure first Omni co vs sep and Directional co vs sep].
  2. THEN I need to know the mean differences between groups (omni and directional). So, I need to compare the mean differences of 0, +4, +8, and +12 for each speaker orientation across omni-directional and directional groups.
  3. For the above two analyses, refer to Table 2, I need to know if they are significantly different from one another.
  4. From here, I need the Directional Benefit/Spatial Advantage which is just the spatially separated conditions minus the collocated condition (Table 3 it is already calculated).
    1. Need to determine if the directional group is significantly more beneficial than the omnidirectional group.
  5. Then I need to perform predictions using probably linear regression.
    1. Specific Aim 1: Can ANL scores or QuickSIN scores predict the Directional Benefit/Spatial Advantage for any of the SNRs (ANL and QuickSIN in Table 2, Spatial Advantage shown in Table 3)
    2. Specific Aim 2: This is more of an exploratory analysis so I am not highly concerned with my low n and lots of measures. Can any subjective measure (APHAB unaided, APHAB aided, APHAB benefit, SSQ Speech, SSQ Spatial, SSQ Qualities, or SIG) predict Spatial Advantage (shown in Table 3), QuickSIN, or ANL (shown in Table 2).
  6. To note: I would like to know the r values to know the effect sizes

I attached my Excel Document for the data as well. I am also attaching my R Code for my attempts. The coding equations may or may not be correct but I do not understand the results either.

Everett Data Analysis

Everett R Code

Initial Meeting
I. Who:
Client: Alyssa Everett, AuD, CCC-A from the Department of Speech, Language, and Hearing Sciences
Consultants: Dean, Yuankun(author), Wen-Wei and Hanh
II. When:
03 October 2019, 1-2pm
III. Discussion:
Our group met with Alyssa Everett, AuD, CCC-A from the Department of Speech, Language, and Hearing Sciences, where she presented us with the information for her study. At the time of the meeting, she has finished data collection with a few missing values, she also had clear questions that she tries to answer from the data. She would like us to deliver the results before her presentation in November. The detail of the study is summarized below:
People that who are over 65 years old are invited to the study for hearing tests. Once they arrive at the facility, first they will be randomly assigned one of the two types of hearing aid, omni-directional or directional. Omni-directional collects sound from all directions, while directional collects only from the front. Then another testing condition will be randomly assigned for the booth, whether the signal and noise are collocated, or spatially separated. Finally they will be given a hearing test at four different signal to noise levels (SNR), 0, 4, 8 and 12. The numbers mean the difference in dB between signal and noise, for example, 12 means signal is 12 dB higher than the noise, the four levels of SNR are assigned in random order. After the test, a score from 0 to 25 will be received by the test taker. The scores can be further complied into a measure called “CRM”.
The Coordinate Response Measure (CRM) is calculated as follows: for each participant, within the “omni-directional” hearing aid, at each of the four levels of SNR, take the score for “separated”, minus the “collocated” scores, those four differences can be seen as the spatial advantage between the two test conditions. The same is then calculated for “directional” hearing aids. In the end there are 8 values recorded and will be used for later analysis.
For each participant, a second set of test will be performed, acceptance of noise test (ANL) and speech-in-noise test (QuickSIN) were given without the hearing aid. The ANL tests a participant’s tolerance to noise while listening to a speech and is more of a self-reported score. QuickSIN is a test where the participant is given a sentence with background noise and asked to repeat the sentence, then a score is given based on how many words he or she repeats correctly.
For a single participant, all the tests were performed on the same day, although different participants could receive tests on different days.
The goal of this study is to predict hearing aid success and determining which microphone mode to use prior to the fitting. The specific tasks are the following:
(1) Within each of the hearing aid (omni / directional), determine the mean differences of collocated and spatially separated conditions, at the four  levels of SNR.
(2) Determine the mean difference of the scores in omni/directional, at the four levels of SNR.
(3) Determine if the above means are significantly different from on another.
(4) Determine the CRM, and find out whether the ANL/QuickSIN scores can be a predictor for the CRM.
Task (1), (2) and (3) requires analysis on the dataset that all independent variables are factors, as suggested by Prof. Billheimer we should attempt to fit a linear mixed effects model (can be accessed from package (nlme) from R).
Task (4) could be answered by fitting an regression of CRM vs ANL/QuickSIN.