Related Experiment Videos
Estimating intraclass correlation for binary data.
M S Ridout1, C G Demétrio, D Firth
1Horticulture Research International, West Malling, Kent, UK. martin.ridout@hri.ac.uk
Biometrics
|April 25, 2001
Summary
This study compares various intraclass correlation estimators for binary data. Several useful methods were identified, but extended quasi-likelihood estimators showed bias in certain situations.
Area of Science:
- Biostatistics
- Statistical Methods
Background:
- Intraclass correlation is crucial for analyzing clustered or repeated binary data.
- Numerous statistical estimators exist for binary data, but their performance varies.
Purpose of the Study:
- To review and compare different intraclass correlation estimators for binary data.
- To identify reliable estimators through simulation and highlight potential biases.
Main Methods:
- Extensive simulation study evaluating various intraclass correlation estimators.
- Comparison of specific estimators and those derived from general methods like pseudo-likelihood and extended quasi-likelihood.
Main Results:
- Several intraclass correlation estimators demonstrated utility for binary data.
- One previously unconsidered estimator showed promise for binary data applications.
- Extended quasi-likelihood estimators exhibited significant bias under specific conditions.
Conclusions:
- The choice of intraclass correlation estimator for binary data is critical.
- Simulation studies are essential for validating estimator performance and identifying biases.
- Researchers should exercise caution with extended quasi-likelihood methods for binary data due to potential bias.