My research is focused on building better predictive models through improved estimation of regression parameters. I am interested in the development and application of James-Stein type shrinkage estimation in parametric and semiparametric regression models and their comparisons with penalty estimators. In the event of a trade-off between bias and prediction error, James-Stein-type methods tend to achieve reduced overall prediction error that offsets the bias in the estimation.
Other areas of interest include
- data mining and machine learning
- design of experiments
- longitudinal data analysis
- Bayesian and MCMC methods
- missing data analysis
- application of statistical methods in education, environment, public health and epidemiology