Statistics 578: Regression Analysis
- Course information, assessment procedures,
etc. See also the legal stuff to
be included with outlines.
of lectures, assignments and exams.
notes. These are constantly under revision.
- The course
assumes an undergraduate knowledge of Regression at the level of STAT 378,
and mathematics at the level of STAT
512. If you are not currently taking 512, you should at least work
through the 512
lecture notes – especially the parts on linear algebra and matrix
theory (including matrix differentiation), and asymptotic distribution
theory. A summary of the required material is in this appendix.
and using the R package:
a quick start you can obtain R here and
install it on a PC by double-clicking on this .exe file and following the
instructions. There is a Mac version of R too; you'll need to consult the
'R home page' link below for this.
- What is R?
- The R home page
- R manuals: small, big.
- The importance of
clarity of exposition, and of grammatical correctness, in technical
writing cannot be over-emphasized. The best way to determine if you
understand what you are doing is to try to write it down in a form that
another reader can understand. Some comments along these lines are
included with the course information; for
some helpful resources see "Writing Aids" on my homepage.
R examples worked in class: You should
find it useful to download and run these R scripts before the classes in
which they are to be discussed.
- cars (Lecture 1) Some preliminary regression
- acetylene1 (Lecture 4) and acetylene 2 (Lecture 5) An analysis of
a dataset exhibiting extreme multicollinearity; ridge regression
- logistic (Lecture 6) Example of logistic
regression through wls
- outlier (Lecture 7) Simulated example of
the effect of one highly influential outlier
(Lecture 8) Fitting a Michaelis-Menten model to the puromycin
- bod1 (Lecture 9) Likelihood regions and
profile plots for the BOD data
- lubricant (Lecture 10) Some methods of
obtaining starting values
- smoothing1 (Lecture 12) Smoothing
spline (and other) fits to the motorcycle data
- smoothing2 (Lecture 13) Kernel and
- rock (Lecture 14) GAM and projection pusuit fitting on the 'rock' data set
(Lecture 14) Simulated projection pursuit example
- lasso (Lecture 15) Computing the lasso
- quantile regression (Lecture 15); see
- ordinary M-estimation
(Lectures 17, 18) Computing an Ordinary M-estimate with Huber's psi
- mineral (Lectures 20 -- 23) Various
regression fits to the 'mineral' dataset from Maronna,
Martin & Yohai
- my GM estimator (Lecture 22)
My function to compute GM-estimates
Matlab code for robustness
sets used in assignments:
- bod2 (Assignment 2)
- steady state (Assignment 2)
(Assignment 3): download water.dat and
water.R to the
same directory, and run water.R. Check
that all 20 cases have appeared.
Statistics resources at U of A:
of Doug Wiens
of Mathematical and Statistical Sciences Home Page