STAT 832: Multivariate Statistical Analysis
- Instructor: Peter Hoff
- TA: Michael Jauch
- Lecture: MW 10:05-11:20, Perkins LINK 087
- Office hours: TTh 10:30-11:30 Rm 219 (PH), T 2:30-4:30 Rm 211A (MJ)
- Lecture notes, code and data
- Applied Multivariate Statistical Analysis (Hardle and Simar)
- Modern Multivariate Statistical Techniques (Izenman)
- Sakai link
2017-04-12: Do the twelfth and final set of exercises, exercises 1,2a and 3 in the multilinear model notes, to be turned in Wednesday 2017-04-19.
2017-04-05: Do the eleventh set of exercises, exercises 1,2,4 and 5 in the multivariate regression notes, to be turned in Wednesday 2017-04-12.
2017-03-29: Do the tenth set of exercises,
exercises 6,7 and 8 in the copula notes,
to be turned in Wednesday 2017-04-05.
2017-03-22: Do the ninth set of exercises to be turned in Wednesday 2017-03-29.
2017-03-01: Do the eighth set of exercises to be turned in Wednesday 2017-03-08.
2017-02-22: Do the seventh set of exercises to be turned in Wednesday 2017-03-01.
2017-02-15: Do the sixth set of exercises to be turned in Wednesday 2017-02-22.
2017-02-08: Do the fifth set of exercises to be turned in Wednesday 2017-02-15.
2017-02-01: Do the fourth set of exercises to be turned in Wednesday 2017-02-08.
2017-01-25: Do the third set of exercises to be turned in Wednesday 2017-02-01.
2017-01-11: Do exercises 1 through 7 in the course notes on eigendcompositions, to be turned in Wednesday 2017-01-18.
Tentative schedule of topics
- Matrix decompositions and linear methods
- Nonlinear methods
- Multivariate normal theory
- General linear model
- Equivariant estimation and testing
- Distributional results for linear methods
- Shrinkage estimation and Bayesian methods
- Tensor data
- Tensor decompositions
- Multilinear regression and covariance models
- Copula models
- Parametric methods
- Semiparametric methods
- Distributions over special manifolds
- Directional data
- Models for eigenvectors
- Big data
- Graphical models and sparse estimation
- Optimal shrinkage when $p\approx n$
- 60% Homework
- 20% Data project
- 20% Paper project
- grade = x exp(-l/8), where x=score, l=days late (including weekends)