Syllabus
Announcements:
Our lecture schedule and location have changed. We now meet
Mon and Wed from 4.45 to 6.00pm, in 104 Thomas.
Office hours have changed, too:
Francesca Chiaromonte: Mon, Wed 3.30--4.30pm, and by appointment.
Thomas 411.
Jiping Wang: Tue, Thur 1.00--2.30pm. Thomas 301.
Francesca's office hour on Mon Oct 16: 2.00pm--3.30pm approximately.
Lectures:
Homework Assignments:
Assignment#1: Practice: Read Chapters 1 and 2 of Cook and
Weisberg
Problem 1.4
Grading: Problems 1.2 and 2.1 (due in class on Wed Sept 6)
Assignment#2 : Practice: Read Chapters 3 and 4 of Cook and Weisberg
Problems 3.2, 4.6
Grading: Problems 3.1, 4.1, 4.2, 4.3, 4.4, 4.7 (due in class on Wed Sept
20)
Assignment#3: Practice: Read Chapter 5 and Chapter 6 up to page
110
plus complements 6.7.1 and 6.7.2
Go through demo on probability plots
Problems 5.4, 6.2, 6.12
Grading: Problems 6.6. (6.6.1 only), 6.9, 6.14, 6.15 (due in class
Mon Oct 2)
Assignment#4: Practice: Finish reading Chapter 6, including complements
6.7.3, 4, 5, 6.
Problems 6.4, 6.8, 6.10
Grading: Problems 6.6 (6.6.2 and 6.6.3), 6.7 (all), 6.13 (due in class
Wed Oct 11)
Assignment#5: Practice: Read Chapter 7 up to 7.5 included (p.158), and
the complements
7.9.1--7.9.5. Review linear algebra concepts if needed. Start reading
Chapter 8 (on using 3D plots) on your own.
Problems 7.2, 7.6
Grading: Problems 7.1, 7.3 (7.3.1--7.3.4 only), 7.4 (due in class Wed Nov
1)
Assignment#6: Practice: Finish reading Chapter 7 and Chapter 8.
Problems: 7.9 (all), 7.10, 8.6
Grading: Problems 7.3 (7.3.5--7.3.8), 7.7 (all), 8.7 (due in class Mon
Nov 13)
Assignment#7: Practice: Read Chapter 9 and Chapter 14, work with the
caution.lsp file illustrating
how residual plots can be missleading.
Problems: 9.5, 9.7 (nice for graphics), 9.10, 14.1
Grading: Problems 9.6, 9.9, 14.2, 14.3 (all except 14.3.2) (due in class
Mon Nov 20)
Assignment#8: Practice: Read Chapter 15 and Sections 10.5, 10.6, 10.7
(added variable plots)
Problems: 14.5, 15.1, 15.2, 15.3, 10.4, 10.12
Grading: 15.4, 15.5, 15.6, 10.5 (due in class Mon Dec 4)
Assignment#9: Practice: Read Sections 10.1.5 (variance inflation) 10.8
(joint confidence regions)
12.4 (models with continuous and categorical predictors), and Chapter 11
(model comparison and terms selection).
Problems: 10.11, 12.6, 11.5, 11.6, 10.9, 12.3, 11.1
NOTE NONE OF THE PROBLEMS IS FOR GRADING ANYMORE!
Mid-Term tests:
THE FIRST MID-TERM HAS BEEN MOVED TO MON OCT 16
Coverage Chapters 1 to 6 of Cook and Weisberg -- concentrate on Chapters
3, 4 and 6.
Some material from Assignment#4 will be on the test (we will try to
have the hmw's returned
to you by Fri Oct 13).
Practice suggestions:
The following are some general topics you can expect to see on the test,
with related problems from
Cook and Weisberg.
- Using lowess and lowess+/-SD to visualize mean and variance functions
on a scatter plot (e.g. 3.2)
- Properties of the bivariate normal distribution (e.g. 4.1, 4.2, 4.3)
- Using probability plots (e.g. demo-prb.lsp, 5.4)
- What happens to the various outputs of the least square fit of a
simple linear regression
when x and/or y are subject to linear transformations
(e.g. 6.2)
- Least squares optimization for a sub-model of the simple linear regression
model.
Nested comparison (ANOVA table, F ratio, and its relation
to the T ratio of the coefficient
set to 0 in the simple linear regression model) (e.g.
6.6)
- Properties of residuals and fitted values (e.g. 6.15)
- fill the slots of a typical regression output with the corresponding
formulae (e.g. 6.1)
- computing confidence interval for the mean response and prediction
intervals for the
response, relations between their width at given levels.
THE SECOND MID-TERM WILL TAKE PLACE ON WED NOV 29
Coverage, Chapters 7, 9, 14, 15.
Practice suggestions:
Multiple linear regression models (Ch 7):
Interpretation of terms (see pg 146, 147), confidence and prediction
intervals, tests for
coefficients and tests for comparison of nested models. Go over problems
7.4, 7.7, 7.8.
The geometry of linear models, e.g. orthogonality of residuals and
terms and of residuals
and fitted values, interpretation of the detrmination coefficient as
the squared sample
correlation between observed responses and fitted values (problem 7.6),
etc.
Diagnosing and remeding misspecifications of mean and variance function
(Ch 9, 14):
The use of residual plots for spotting mean and variance misspecification.
The formal
tests, which include curvature test and score test for non constant
variance (Ch 14),
as well as lack of fit test procedures (Ch 9). Beyond constant variance:
variance modeling
and weighted least squares (Ch 9). Go over problems 9.1, 9.4.(1,2,4,5),
9.11, 14.1, 14.2.
Case Diagnostics (Ch 15):
Characterizing cases in terms of leverage (Ch 7), residuals/studentized
residuals, and
influence (Ch 15) -- points that are "separated" from the bulck in
terms of u's, points
that are "separated" from the bulk in terms of the relationship between
response and
u's. The issue of masking. Testing for mean-shift outliers.
You haven't yet had a chance to hand in and get back homework on this
part. Some
questions on it will be in the test, although they will carry marginal
weight. To prepare,
go over the practice and grading problems in Assignment#8 above (not
the ones relative
to AVP's)
Final Project:
Due by 12.00 noon on Thur Dec 14 (to me, or the Statistics main office
-- do not use my
mailbox, to avoid that projects get lost).
Data set:
in format digestible by ARC with a text extension final_d.txt
in format digestible by ARC with a xlispstat extension final_d_ARC.lsp
as an excell sheet (which can be loaded in MINITAB) final_d_EXCELL.xls
(let me know asap if you have trouble unloading the data)