CSI 991
Seminar in Computational Statistics:
New (and Old) Ways of Looking at Regression
Spring, 2003
Fridays 3:00pm -- 5:00pm, Enterprise, Room 175
Contacts:
csutton@gmu.edu
jgentle@gmu.edu
We will work through some of the material in
Regression Modeling Strategies
With Applications to Linear Models, Logistic Regression, and Survival Analysis
by Frank Harrell, beginning with Chapter 7.
Website for text.
We will also do some exploratory analyses in S-Plus, CART and MARS.
During the first couple of weeks, we'll finish up some loose ends from
the fall seminar, i.e., material in the first 6 chapters of Harrell.
See discussion of projection matrices
for example.
Schedule
-
January 17 (meet in ST I, Room 206)
Discuss plans for the semester.
-
January 24
Experiment with the "counties" dataset of Harrell, using his
S-Plus code in Chapter 7
(thanks to James Walter!), and using regression trees in either S-Plus
or MARS.
*** There is a problem with the %ia% used in the code at the bottom
of page 128. Carlos will check with Harrell about possible fixes.
-
January 31
Discuss Spearman's rho^2.
If the data follow an exact quadratic relationship,
Spearman's rho^2 is 1 if the inflection point is outside the range of
the data, but is less than 1 if the inflection point is inside the range of
the data.
Questions for thought:
(Important) In the case of a quadratic relationship with the inflection point
inside the range of the data, will Spearman's rho^2 be uniformly larger than
Spearman's rho?
(Irrelevant, but interesting) In the case of equally-spaced data following
an exact quadratic relationship, how does Spearman's rho^2 vary as the
inflection point moves from one end of the range of the data to the other?
Conjecture: It begins at 1, goes to a minimum around the 25% quantile, increases
to a local maximum (less than 1) at the 50% quantile, goes to a minimum
around the 75% quantile and then increases to 1 at the other end of the range.
-
February 7
-
February 14
-
February 21
-
February 28
-
March 7
-
March 14
-
March 21
-
March 28
Notes on GLIMs.
Data in example (cancer.dat).
-
April 4
-
April 11
-
April 18
-
April 25
Mark's Chapter 11 code.
-
May 2 -- May 30
Survival analysis (Chapters 16 -- 19)
Introductory material.