Related Experiment Videos
Ordinal regression model and the linear regression model were superior to the logistic regression models.
Colleen M Norris1, William A Ghali, L Duncan Saunders
1Faculty of Nursing, University of Alberta, Edmonton, Alberta, Canada. colleen.norris@ualberta.ca
Journal of Clinical Epidemiology
|April 25, 2006
Summary
For skewed health-related quality of life data, linear and ordinal regression models offer more stable estimates than logistic regression. Combining both provides the most thorough data interpretation.
Area of Science:
- Biostatistics
- Health Outcomes Research
Background:
- Ordinal scales frequently yield skewed data, complicating analysis.
- The best statistical approach for skewed health-related quality of life (HRQOL) data remains unclear.
Purpose of the Study:
- To compare four multivariable statistical strategies for analyzing skewed HRQOL data.
- To evaluate the performance of linear, logistic, and ordinal regression models.
Main Methods:
- Methodological study comparing four regression models: linear, two logistic, and ordinal.
- Analysis involved assessing odds ratios, confidence intervals, and confidence interval widths.
Main Results:
- Linear and ordinal regression models demonstrated more stable parameter estimates.
- These models also exhibited narrower confidence interval widths compared to logistic regression.
Conclusions:
- Linear and ordinal regression are more robust for skewed HRQOL data.
- A combined analysis using both adjusted scores and odds ratios offers the most comprehensive interpretation.