|
Activity Number:
|
342
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Wednesday, August 6, 2003 : 9:00 AM to 10:50 AM
|
|
Sponsor:
|
Section on Health Policy Statistics
|
| Abstract - #300024 |
|
Title:
|
Likelihood Modeling of Visual Acuity Norms
|
|
Author(s):
|
Angie Mary Wade*+
|
|
Companies:
|
Institute of Child Health, London
|
|
Address:
|
Centre for Epidemiology and Biostatistic, London, Intern, WC1N 1EH, England
|
|
Keywords:
|
Age-related normal ranges ; Ordinal ; Maximum-likelihood estimation
|
|
Abstract:
|
Visual acuity is measured via letter charts on an ordinal scale. Detection of defective vision in children is complicated by the natural development of acuity from birth. Whilst there is a wealth of literature on the estimation of age-related normal ranges for continuous outcomes, very little has addressed the particular analytic problems when the outcome is ordinal. We have developed a method using proportional odds models and maximum likelihood estimation of model parameters to create smoothly changing normal standards for visual acuity. Asymmetric logistic models are shown to be ideal for data of this form. The approach used has several positive advantages over alternative methods. The technique allows formal investigation of subgroup differences (levels or patterns of acuity may vary between boys and girls, for instance) via likelihood ratio tests and estimation of confidence intervals via profile likelihood. Model parameters may be smoothed across categories to facilitate the estimation of sd scores for individual children but this is shown to be inappropriate in our dataset. The methodology is illustrated using measurements collected from 3,000 children between two and nine years of age.
|
