|
Activity Number:
|
302
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Wednesday, August 11, 2004 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
WNAR
|
| Abstract - #300349 |
|
Title:
|
Predicting Hearing Threshold in Nonresponsive Subjects Using a Bayesian Linear Model in the Presence of Left-censored Covariates
|
|
Author(s):
|
Byron J. Gajewski*+ and Judy Widen and Nannette Nicholson
|
|
Companies:
|
University of Kansas Medical Center and University of Kansas Medical Center and University of Kansas Medical Center
|
|
Address:
|
3901 Rainbow Blvd., Kansas City, KS, 66160,
|
|
Keywords:
|
lognormal ; location shift
|
|
Abstract:
|
This paper considers the prediction of pure tone thresholds for populations with the inability to respond to usual hearing tests. This is important because early intervention of hearing problems in infants, for example, is imperative for future development and quality of life. A standard statistical method in the hearing literature is to define a level of adequate hearing threshold and utilize categorical techniques such as discriminate analysis or logistic regression to diagnose based on a reproduction of a signal sent through the ear that does not require a subject's active response. The problem is that the signal suffers from censoring due to measurement constraints. Thus, the current methods deflate the leverage in the prediction model. We remedy this shortcoming by modeling the independent variables with a normal distribution while accounting for the censoring. Additionally, expanding from a previous hearing threshold model, we provide the audiologist a prediction interval of the actual pure tone threshold rather than a dichotomous prediction of hearing deficiency.
|
