Related Experiment Videos
On the consequences of model misspecification in logistic regression
1Division of Biostatistics, Columbia University, School of Public Health, New York, NY 10032.
Environmental Health Perspectives
|July 1, 1990
Summary
Model misspecification in logistic regression can impact association tests. This review examines how errors like measurement issues or omitted variables affect likelihood score tests for exposure-response relationships.
Area of Science:
- Biostatistics
- Epidemiology
- Statistical Modeling
Background:
- Logistic regression models are standard for binary outcomes.
- Controlling for covariates is crucial in these models.
- Model misspecification can compromise study findings.
Purpose of the Study:
- To review consequences of model misspecification in logistic regression.
- To analyze effects on likelihood score tests for exposure-response association.
- To cover various misspecification types, including measurement errors and omitted variables.
Main Methods:
- Review of recent statistical research.
- Analysis of large sample properties of likelihood score tests.
- Focus on association tests in logistic regression.
Main Results:
- Model misspecification adversely affects tests of exposure-response association.
- Errors in exposure or covariates can lead to biased results.
- Consequences depend on the type and extent of misspecification.
Conclusions:
- Careful model specification is vital for valid inference.
- Understanding misspecification effects is key for robust statistical analysis.
- Further research needed on mitigating misspecification impacts.