Related Experiment Videos
A SAS macro for residual deviance of ordinal regression analysis
1Division of Biostatistics and Epidemiology, University of Tennessee, Memphis 38163, USA.
Computer Methods and Programs in Biomedicine
|December 1, 1994
Summary
This study introduces a SAS macro to calculate the saturated model likelihood for ordinal regression analysis. This tool enables goodness-of-fit testing using residual deviance, crucial for accurate statistical modeling.
Area of Science:
- Statistics
- Biostatistics
- Computational Statistics
Background:
- Ordinal regression models are used for outcomes on an ordered scale.
- Assessing model fit is critical for reliable statistical inference.
- Standard SAS procedures lack direct output for residual deviance in ordinal regression.
Purpose of the Study:
- To present a SAS macro for calculating the saturated model likelihood in ordinal regression.
- To enable goodness-of-fit testing via residual deviance for ordinal regression models.
- To supplement existing SAS procedures for enhanced statistical analysis.
Main Methods:
- Development of a SAS macro to compute the saturated model likelihood.
- Application of the macro to ordinal regression with multinomial outcomes.
- Utilizing residual deviance and chi-squared distribution for model fit assessment.
Main Results:
- The SAS macro successfully calculates the saturated model likelihood.
- The macro facilitates the computation of residual deviance for goodness-of-fit testing.
- Demonstrated utility with data from an ongoing HIV study.
Conclusions:
- The developed SAS macro provides a valuable tool for ordinal regression analysis.
- It enhances the ability to perform goodness-of-fit tests in SAS.
- This macro aids researchers in validating their ordinal regression models.