University of Minnesota
Course Information for Stat 8053
Fall Semester, 2011
2:30-3:20 MWF (09/07/2011 - 12/14/2011), 206 Vincent

General Information

The instructor is S. Weisberg, 312 Ford Hall, 612-625-8355. I don't have fixed office hours for this course, but will happy to make appointments. Send an email to sandy@stat.umn.edu.

Textbooks

Computing

Computing will use the program R, which is available on the School Linux network, and can also be downloaded for your own computer from www.r-project.org. Familiarity with R from Stat 8051-2 is assumed.

The handout www.stat.umn.edu/~sandy/courses/8051/handouts/usingR/ from Stat8051 fall 2007 might be helpful if you are new to R, or if you would like to install LATEX on your Mac or PC.

Web Site

The web site for this class is http://www.stat.umn.edu/~sandy/courses/8053.

Homework

For each homework assignment, one student will be assigned as the ``Leader" and two as ``Support". This group is responsible for presenting the material covered in the homework; for providing at least partial solutions to the problems; and for reading and correcting the homework of the other students. I will make the assignment of leaders and helpers (using a balanced incomplete block design) and email it to you before Friday of the first week of class.

The Presentation The presentation should be a Beamer/Power Point presentation on the computer. You will also need to prepare a handout that summarizes what you have done. If the handout has computer output, it should either be annotated on the handout, or else discussed in class so the other students can add annotation as you go over the handout. Students who are not presenters may have different solutions to problems and these should be discussed during the presentation.

Presentation Tools You are not required to use Beamer, but if you choose to do so the presentation at www.stat.umn.edu/~sandy/courses/8801/handouts/03.Beamer/beamer.pdf might help you get started. If you would like to use Sweave as well for ``reproducible" handouts, look at www.stat.umn.edu/~charlie/Sweave/.

Grading The group of three for the week can decide among themselves who will grade what. In previous years the leader did all the grading for the week. Each assignment is worth ten points; the grader should email me the grades within a week..

What should be on the presentation Let's take Assignment #1 as an example. The first problem concerns a data problem for which a linear mixed model is probably appropriate. The assigned problem is relatively vague, but the homework assignment gives a hint to the general procedure you should follow. You may need to use several R packages beyond the standard packages, and it is assumed that you have seen some of these in 8051/2. In any case, all data-analysis problems, the bulk of this course, will consist of (1) initial exploration of the data, the understand the basic structure. Occasionally data sets used in homework have gross errors and missing values, and you need to find these. After you are confident about the structure of the data, you can proceed to (2) modeling, but you would want to be able to answer questions like: is it reasonable for this factor to be random, that term to enter the model linearly, and so on. This will become easier as the course goes on. Your presentation will present some R details, lots of graphs, and words that summarize your findings.

The presentation for the second problem should be much like the first. The questions are more general, so you need to decide on which methods/summaries are useful, and what you have learned from the analysis.

The third problem requires you to derive a likelihood function, but not do any data analysis. You will probably want to begin with the likelihood function for a usual GLMM (or maybe for any mixed-model with a linear predictor $\eta = x'\beta + z'u$ with $u$ random). The twist in the problem is that the predictors are unobservable, but rather you only observe sums of the predictors. If you can write down the likelihood function you would then want to give as much detail as you can about how you might maximize it. Again, this was not discussed in class, so you will have to find this information on your own.

If there are more students than homework assignments, some students will never get to be the ``Leader" and present homework. These students will, however, be required to give a presentation on another topic during the second half of the course. Details will follow later.

Working Together

Collaboration on homework is strongly encouraged, but each student must submit their own homework paper. Collaboration on exams is not allowed.

Course outline

This will be a survey course, covering a different topic each week with topics from the Faraway book for the first five weeks and topics from the Härdle and Simar book for the remainder of the course. The usual outline for this course will be: Monday and Wednesday, lectures on the topic-of-the-week, and Friday of the following week, discussion of homework and computing issues. The first homework will be discussed on Monday, Sept. 19).

Class will not meet on Friday, November 4.

Exams and Grading

Homework, including your class presentation, will account for 40% of the grade.

A mid-term exam, covering the material from the Faraway book, will count for 30% of the grade. The final exam, covering the whole course, will also count for 30% of the grade.

Students who fail to complete all work will receive a grade based on the work done. Grades of incomplete will be given only in extraordinary circumstances, and only after negotiation with the instructor.

Collaboration on the exam is absolutely prohibited. Any students who work together on exams will be given a failing grade in the course.

Handouts

Copies of the handouts are available from the class web page.

Disability Access Statement

This publication/material, and all other handouts in Statistics 8053, is available in alternative formats upon request. Please contact School of Statistics, 313 Ford, 625-8046.

Course Outline

Faraway, Chapters 10-14 followed by Härdle and Simar Chapters 4 to 18, plus some newer stuff from the current literature.

Return to Stat 8053 home.


S Weisberg
2011-09-06