Overview
Michael J. Daniels is a biostatistician known for Bayesian methods for covariance matrix estimation, missing data in longitudinal studies, and semiparametric models. He was at Iowa State University at the time of his collaboration with Robert Kass and subsequently moved to the University of Florida.
Key Contributions / Features
- Daniels-Kass (1999) (JASA 94: 1254–1263): Full nonconjugate Bayesian estimation of covariance matrices via MCMC — the computationally intensive predecessor to the 2001 paper.
- Daniels-Kass (2001) (Biometrics 57(4): 1173–1184): Two-stage empirical Bayes shrinkage — log-eigenvalue posterior mean estimator (closed-form, no MCMC), correlation shrinkage, and rotation shrinkage via Givens angles (not recommended in practice); PRIAL up to 70%.
- Daniels (1999) (Canadian Journal of Statistics 27: 569–580): Shrinkage estimators under alternative loss functions.
- Daniels-Cressie (2001) (Journal of Time Series Analysis 22: 253–266): Covariance estimation for spatial-temporal longitudinal models.
- Daniels-Pourahmadi (2001): Semiparametric covariance modeling via modified Cholesky decomposition.
Related