Department of Statistics

STATS 760: Regression Modeling

Course Information 2006

Lecturer:

Alan Lee

Department of Statistics

City Office: Room 228, Mathematics and Physics Building

Tamaki Office: Room 332, Building 721

Telephone : 373 7599 extension 88749 (Tamaki) 88751 (City)

Fax: 373 7000

email: lee@stat.auckland.ac.nz

Course web page: www.stat.auckland.ac.nz/~lee/760

Introduction:

The aim of this course is twofold: to give you a fairly comprehensive survey of modern applied statistics, and give you some training in how to find out details of statistical techniques that are unfamiliar to you. If you are working in statistics after graduation, you will have to teach yourself a lot of new techniques. This course is designed to give you practice in doing this.

Thus, the course is rather different from most postgraduate courses, in that there will be very few lectures, and most of these will be given by you!

The course is based on the book “Modern Applied Statistics with S-Plus” by W.N. Venables and B.D. Ripley. You will work through selected chapters of the book, at your own pace. Each student will be assigned 5 topics, chosen from chapters 3 and 6-14 of Venables and Ripley. There are currently 4 editions of this book, but these chapters in all editions correspond roughly to the following 10 topics:

Plotting and Graphical displays

Linear models

Non-linear regression

Generalized linear models, including GAMS

Modern regression, including smoothing, regression trees and neural nets

Mixed models

Multivariate analysis

Survival analysis

Time series analysis

Spatial statistics

I will assign each student 5 of these topics. Thus, each of you will have a different study program.

Course organization:

I will have a one-on one meeting with each of you once a week, at a time and place to be arranged. This should last about 30 minutes. At the meeting, you will discuss how you are progressing with your mastering of the assigned chapters. You will be expected to research the topic, and gain the following

· An appreciation of why the technique is important

· The kinds of applied problems that the technique can solve

· The types of data that call for the technique

· The software (mainly R but sometimes SAS) that you need to implement the technique

· How to interpret the output from the computer runs

· Any diagnostic techniques that are important

· How the technique relates to others (e.g. how to glms relate to linear models, GAMS to glms)

In addition, you will be assigned one of the 5 topics for more formal study. For this topic, I will require you to write a short paper (around 20 pages) describing the technique and covering the bullet points above. In addition, you will present a 20 minute lecture to the class describing your findings. For the paper and presentation, you will work in pairs.

I will also require you to keep a journal, in which you will make notes recording what you are learning. You will also include computer output in your journal. Discussion of your journal entries and computer runs will form the basis of our meetings. Journals are not formal documents – they are a record of your discovery process, and can be handwritten. However, I will have to be able to read them!!

Assessment:

This will be on the basis of your journal, your paper and presentation, and also on the basis of an oral exam at our final meeting when we will review what you have learnt. The breakdown will be

Journals/meetings, including the final 40%

Oral exam 30%

Paper 20%

Oral presentation 10%

NB: There is no final exam.

Getting started - the first week:

At the first lecture (on Tuesday Feb 28, in Rm 721.231 at Tamaki), I will discuss the assignment of topics. These will be subject to negotiation, but in general I will expect the material covered to be new to you. We will also discuss how you can pair up for the paper and the presentation.

In addition, I will discuss the resources you can use to research your topics. The primary resource will be the textbook, but there are many excellent texts you can use for supplementary reading. There is also the R and S-Plus online help, plus SAS manuals and books. The internet (particularly using Google) can also be very helpful. Copies of the textbook will be made available through the Stats office. A supplementary reading list will also be given out.

In weeks 2-10, I will give the occasional “overview” lecture on the topics that are unfamiliar to most students. These will cover non-linear regression, modern regression, mixed models, survival analysis, time series and spatial statistics. The lectures will be very introductory and will be designed to get you started on these topics – times and dates to be arranged. The rest of the time will be taken up with self study, our weekly meetings, and preparation of your papers and presentations. The presentations will be held in the last week of term. I expect all students to attend all these lectures and be prepared to make constructive comments during question time. Papers will be due on the last Friday of the semester (Friday 2nd June).

Collaboration:

It is my view that collaboration is an important part of the learning process and I encourage you to discuss the material you are studying with each other (and me!) However, you must not copy another person’s journal.

Course Planner:

In the diagram below, 0 is the first week, discussed above. The other figures denote “periods” and are a guide to what topics you should be working on in any particular week.

You will be assigned one topic for each period.

Week
Beginning	Monday	Tuesday	Wednesday	Thursday	Friday
27/02/06	0	Lecture	0	0	0
6/03/06	1	1	1	1	1
13/03/06	1	1	1	1	1
20/03/06	2	2	2	2	2
27/03/06	2	2	2	2	2
3/04/06	3	3	3	3	3
10/04/06	3	3	3	3	3
17/04/06	Mid-semester break/Easter
24/04/06
1/05/06	4	4	4	4	4
8/05/06	4	4	4	4	4
15/05/06	5	5	5	5	5
22/05/06	5	5	5	5	5
29/05/06		Talks		Talks	Paper Due

Web page: The course web page is at www.stat.auckland.ac.nz/~lee/760 and I will post various resources on this page, plus administrative notices. Please check it regularly.