To evaluate the impact of different statistical methods on the accuracy of ophthalmic research by accounting for correlations between a patient's two eyes.
Key Findings:
Treating measurements from both eyes as independent leads to inflated Type-I error rates.
Model performance varies based on whether predictors are measured per subject or per eye.
Interpretation:
Selecting appropriate statistical models that account for interocular correlation is crucial to avoid misleading results in ophthalmic research.
Limitations:
The study is based on simulated data, which may not fully capture real-world complexities.
Conclusion:
Mixed effects models and generalized estimating equations are preferred for eye-level predictors, while averaging may suffice for subject-level predictors. Avoid treating both eyes as independent observations.
