Starting Friday, October 4, at 3:30 PM in Room 206 of Science and
Technology 1 (*not* Sci Tech 2), we will begin discussing Frank Harrell's
   Regression Modeling Strategies: With Applications to Linear Models,
   Logistic Regression, and Survival Analysis.

While this book isn't pushing at the edge of the envelope like Hastie,
Tibshirani, and Friedman, it isn't old-fashioned either.  It seems clear
to me from H,T&F that the newer methods are built on a firm foundation of
older methods, and it may be that a book like Harrell's is just what some
of us need to fill in some gaps and set us up to better appreciate
statistical methods which are close to the frontier.  (To me, data mining
is basically regression, classification, clustering, and estimation of
conditional probabilities associated with categorical variables, with an
emphasis on computer-intensive procedures to deal better with large data
sets.  From H,T&F one can see that simple issues such as bias-variance
tradeoff, matrix manipulations, and techniques for variable selection and
model building are important in the "new statistics", just as they are for
the old least squares dinosaurs.  Harrell's book ought to help us get more
comfortable with some of the basic issues, and see connections between
older methods and newer methods.)  Since Harrell's book makes use of
S-Plus, the seminar ought to help some of us learn how to use that
software better.

For the first meeting, we only have 90 minutes.  (Most weeks we will go
from 3 to 5 PM, but on the first Friday of Oct., Nov., and Dec., we will
start at 3:30 PM.)  Let's plan to have read the Preface, Ch. 1 & Ch. 2
prior to the meeting.  This will be more pages than will we typically
cover, but my guess is that the first 20 pages or so may be relatively
simple compared to the rest of the book, and we won't need to spend a lot
of time on them.  I believe (not sure, since my book was only shipped out
from Amazon.com this past Thursday) that the end of Ch. 2 touches on
splines and trees, so there is a possibility that even in Ch. 2 there is
coverage of topics that some of us haven't mastered.

The most important thing will be for you to do the reading each week, and
prepare a short list of questions that you have or topics that you'd like
to discuss.  Since there may be as many as 10 of us this semester, I think
it will be best if we try to have it so just one person is talking at a
time.  (If too many conversations are going on at the same time, it's hard
for people to follow.)

Perhaps a good plan will be that at the start of each meeting, we'll go
around the table and have everyone check in with where they stand with
regard to the material.  (E.g., someone might say, "I've read through
p. 35, but became less confident in my understanding about p. 28 or so."
Another person might warn us that he has a bunch of questions about a
particular subsection, and maybe from time to time someone will say that
she has prepared a 10 minutes presentation on an example from the book,
including the exploration of a possibility that the author overlooked.

CDS