¶¡ÏãÔ°AV

This course is designed to give students some practical experience as a statistical consultant through classroom discussion of issues in consulting and participation in the department's Statistical Consulting Service under the direction of faculty members or the director.

Students will participate in the department's Statistical Consulting Service under the direction of faculty members or the director.

The statistical theory that supports modern statistical methodologies. Distribution theory, methods for construction of tests, estimators, and confidence intervals with special attention to likelihood and Bayesian methods. Properties of the procedures including large sample theory will be considered. Consistency and asymptotic normality for maximum likelihood and related methods (e.g., estimating equations, quasi-likelihood), as well as hypothesis testing and p-values. Additional topics may include: nonparametric models, the bootstrap, causal inference, and simulation. Prerequisite: STAT 450 or permission of the instructor. Students with credit for STAT 801 may not take this course for further credit.

A modern approach to normal theory for general linear models including models with random effects and "messy" data. Topics include experimental units, blocking, theory of quadratic forms, linear contrasts, analysis of covariance, heterogeneous variances, factorial treatment structures, means comparisons, missing data, multi-unit designs, pseudoreplication, repeated measures mixed model formulation and estimation and inference. Prerequisite: STAT 350 or equivalent.

The theory and application of statistical methodology for analyzing non-normal responses. Special emphasis on contingency tables, logistic regression, and log-linear models. Other topics can include mixed-effects models and models for overdispersed data. Prerequisite: STAT 830 and STAT 850 or permission of instructor.

An advanced treatment of modern methods of multivariate statistics and non-parametric regression. Topics may include: (1) dimension reduction techniques such as principal component analysis, multidimensional scaling and related extensions; (2) classification and clustering methods; (3) modern regression techniques such as generalized additive models, Gaussian process regression and splines. Prerequisite: STAT 830 and STAT 853 or permission of instructor.

An introduction to computational methods in applied statistics. Topics can include: the bootstrap, Markov Chain Monte Carlo, EM algorithm, as well as optimization and matrix decompositions. Statistical applications will include frequentist and Bayesian model estimation, as well as inference for complex models. The theoretical motivation and application of computational methods will be addressed. Prerequisite: STAT 830 or equivalent or permission of instructor.