STAT 851 Generalized Linear Models and ... - People.stat.sfu.ca
STAT 851 Generalized Linear Models and ... - People.stat.sfu.ca
STAT 851 Generalized Linear Models and ... - People.stat.sfu.ca
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
<strong>STAT</strong> <strong>851</strong><br />
<strong>Generalized</strong> <strong>Linear</strong> <strong>Models</strong> <strong>and</strong> Discrete Data Analysis<br />
Spring, 2012<br />
Lectures: MW 10:30-12:20 (AQ5118)<br />
Instructor: Rachel Altman<br />
Email: rachelm@<strong>sfu</strong>.<strong>ca</strong><br />
Phone: 778.782.3288<br />
Office hours: MW 1-2pm, or by appointment<br />
Lo<strong>ca</strong>tion: SC-K10551<br />
Course website: http://people.<strong>stat</strong>.<strong>sfu</strong>.<strong>ca</strong>/∼rachelm/<strong>stat</strong><strong>851</strong>.html<br />
(Some) class notes, data sets, S-PLUS/R code, <strong>and</strong> assignments will be available on-line. It<br />
is recommended that you print out the notes before each lecture.<br />
Recommended Textbooks:<br />
1. Categori<strong>ca</strong>l Data Analysis, 2nd Edition, by Alan Agresti. Publisher: Wiley.<br />
2. <strong>Generalized</strong> <strong>Linear</strong> <strong>Models</strong>, 2nd Edition, by P. McCullagh <strong>and</strong> J.A. Nelder. Publisher:<br />
CRC Press.<br />
3. Extending the <strong>Linear</strong> Model with R: <strong>Generalized</strong> <strong>Linear</strong>, Mixed Effects <strong>and</strong> Nonparametric<br />
Regression <strong>Models</strong>, by Julian Faraway. Publisher: CRC Press.<br />
If you are intending on buying only one textbook, I would suggest Agresti. I will be following<br />
the outline of that book, more or less. McCullagh <strong>and</strong> Nelder covers the same material as<br />
Agresti, for the most part, but would be useful if you’re seeking a different perspective or<br />
additional examples. I will be providing substantial sample S-PLUS/R code, but Faraway<br />
might be helpful to those of you who are especially interested in computing.<br />
Computing:<br />
You are welcome to use any software package you’d like to do the data analyses. However,<br />
I will provide support only for S-PLUS/R.<br />
Marking Scheme:<br />
Worth (Tentative) Due Dates<br />
Assignment 1 TBA Jan. 13, 2012<br />
Assignment 2 TBA Jan. 25, 2012<br />
Assignment 3 TBA Feb. 8, 2012<br />
Assignment 4 TBA Feb. 22, 2012<br />
Midterm 20% Mar. 5, 2012 (in class)<br />
Assignment 5 TBA Mar. 20, 2012<br />
Assignment 6 TBA Apr. 5, 2012<br />
Final ∗ 30% Apr. 13, 2012 (9:30am-12:30pm)<br />
1
NOTES:<br />
• ∗ You must achieve 50% on the final exam in order to pass the course.<br />
• Time permitting, you will also be expected to make a short, in-class presentation.<br />
Topics Covered (Time Permitting):<br />
1. Introduction to <strong>ca</strong>tegori<strong>ca</strong>l data<br />
2. Review<br />
• Inference in the univariate <strong>and</strong> bivariate <strong>ca</strong>ses<br />
• <strong>Linear</strong> models, least-squares, matrix notation<br />
• Maximum likelihood theory<br />
3. Theory of generalized linear models<br />
• Model components<br />
– Exponential family <strong>and</strong> its properties<br />
– Link functions<br />
• Maximum likelihood estimation<br />
– Newton-Raphson method<br />
– Iteratively reweighted least-squares<br />
• Goodness-of-fit<br />
– Analysis of deviance<br />
– Pearson <strong>stat</strong>istic<br />
– Analysis of residuals<br />
• Model selection<br />
4. Particular models<br />
• Binary data<br />
• Categori<strong>ca</strong>l data<br />
• Poisson data<br />
• Multinomial data<br />
5. Overdispersion, quasi-likelihood, <strong>and</strong> estimating equations<br />
6. R<strong>and</strong>om effects: generalized linear mixed models<br />
2
Some Final Notes. . .<br />
1. You are expected to be able to communi<strong>ca</strong>te <strong>stat</strong>isti<strong>ca</strong>l concepts both mathemati<strong>ca</strong>lly<br />
<strong>and</strong> in English. You will be marked on your clarity in both contexts.<br />
2. For assignments, please submit only computer output that is relevant to the solution.<br />
(Most software packages provide much more output than you need!)<br />
3. You may discuss assignment problems with your classmates, but work you h<strong>and</strong> in<br />
must be your own.<br />
4. The Code of A<strong>ca</strong>demic Honesty will be enforced:<br />
http://www.<strong>sfu</strong>.<strong>ca</strong>/dean-gradstudies/current/research/a<strong>ca</strong>demic-honesty.html<br />
5. Please come to office hours or email me (rather than dropping by my office) if you have<br />
questions.<br />
6. In general, late assignments are not accepted except in the <strong>ca</strong>se of (documented) illness,<br />
family emergency, etc. If you are having problems finishing an assignment, underst<strong>and</strong>ing<br />
the material before an exam, etc., please contact me well in advance.<br />
3