About this document
Inverse-Wishart Prior Limitations by vankrijg is a document available to read on EtoBox.
The article discusses the limitations of using inverse-Wishart priors in Bayesian methods for random effects models, particularly their tendency to create a dependency between variance and correlation. This can lead to exaggerated correlation estimates when variance is higher than expected, which undermines the desired shrinkage behavior. The author suggests exploring alternative priors or adjusting the inverse-Wishart prior to mitigate these issues.
- Author
- vankrijg
- Language
- EN