Announcements
- August 1, 2002 (1:16 AM)
- I added an
outline of what the course covered.
I could add more to this website (in particular, updating the reading
guide), but it seems rather silly to do so at this point, and besides I
need to focus my attention on some work I should do for the statistical
consulting
center, and on the upcoming fall semester (for which I'll have to craft
a new version of the nonparametrics course, since I'm having to change
books (and I haven't taught the course in 3 years)).
- July 26, 2002 (9:41 PM)
- Some students requested that I post answers to the last few
homework problems, and I have now done so on the
homework solutions web page.
- July 25, 2002 (11:20 PM)
- Well, for some of you, it's Christmas in July --- I decided to
boost the bottom five finishers one letter grade increment (say, B to
B+) from what I had planned to do yesterday. I'm not happy with the
poor performance on the exam, nor the homework scores, for the bottom
bunch of the class, but I decided to adjust for the fact that the
workload and level of difficulty were greater than I had originally
planned.
- July 24, 2002 (11:40 PM)
- I finished grading last night, and nearly have the final letter
grades determined, but I think I'll give the matter a bit more
consideration, and then submit grades sometime Thursday afternoon.
(Since we now submit grades electronically, I think you can get into the
system and check your grades very soon after I submit them.)
The median on the final exam is 59.4%, which a bit bit lower than my
Tuesday afternoon guess of 65%. Six students did really excellent,
ranging from 86.4% to 96.8%, but five scored less than 50%, which is
disappointing. For overall course grades, there will be 7 grades of A,
and 1 A-; but except for those 8 grades, I will wait until tomorrow
before making decisions about the letter grades. Unfortunately, some of
the students who didn't do well on the exam, also didn't do as well on
the homework problems as I would have liked to be the case (largely
because not enough problems were submitted after the July 4th break),
and so the distribution of overall course averages has a strong bimodal
appearance, making it hard to justify too many grades of A- and B+.
Still, given that graduate courses that meet twice a week in the summer
are very tough for many students to manage, anytime that the
"class GPA" is at least 3.5 (which will be the case), I feel that as a
group, the class learned a lot.
I plan to add some more material to the course web site later, but I'll
wrap up the loose ends from this course by Monday morning so that I can
focus on other matters (like the upcoming fall semester, and some work I
need to do for the university statistical consulting center).
- July 23, 2002 (3:31 AM)
-
I added to the
homework solutions web page,
and posted
answers for Quiz #8, and
answers for Quiz #9.
Note that even though the course officially ends with the conclusion of
the final exam, I'll continue to add to this web site for a while,
providing information about grades for the class, and perhaps posting
more homework solutions.
- July 22, 2002 (12:54 AM)
-
I edited the
information about the final exam
web page. I don't plan to add any more hints about what to study.
- July 21, 2002 (2:48 AM)
-
I edited the
information about the final exam
web page. You can note that 14 of the 20 small problems/questions
deal with just 3 relatively small portions of the course material.
- July 19, 2002 (9:54 AM)
-
I added to the
information about the final exam
web page. (Note: Later today, people from the university
computing support staff are planning to switch my e-mail system.
Based on what I've heard from others who have already been switched
over, I anticipate some problems, and it may be the case that I won't be
able to use e-mail over the weekend. But I do plan to be over in the
Central Module for my extra office hours on Sunday and Monday.)
- July 18, 2002 (12:49 AM)
-
I added a few more
corrections to the class notes,
and I added a bit more
information about the final exam.
(I still intend to add a lot more information about the final.)
- July 17, 2002 (6:23 PM)
-
I've posted
answers for Quiz #13, and also
information about your final quiz
averages. In addition, at the very bottom the the home page for
this course website, I've added links to information about R and
S-PLUS, both of which students can get for free.
- July 16, 2002 (1:38 AM)
-
I created a
web page pertaining to the final
exam. I'll add to it over the next 6 days. Please note that it
states that I'll
have extra office hours on Sunday and Monday, but I won't have any this
Thursday.
- July 15, 2002 (6:23 PM)
-
I posted some
hints and comments for Problem 16,
because I'm a bit worried that some of you may mess it up otherwise.
Also, I finally posted the answers to Problem 14 on the
homework solutions web page. If
you desire more details about Problem 14, then maybe it'll be best to
just track me down. While I do hope to provide you with more details
concerning solutions to the homework problems eventually, tonight I'm going to turn
my attention to making up the final exam, and making a web page to
provide you with some guidelines for studying for it.
- July 14, 2002 (4:35 PM)
-
I edited my
hints and comments for Problem 21
(with the new comment, a "be careful" warning of sorts, in color). I'm
in the process of writing about 8 pages of notes on cross-validation and
other methods of model selection and assessment to close the lecture
presentations with. Try to read over the material I gave you on robust
regression and my study comparing regression methods, since I might try
to present that stuff in about 60 minutes or so, in order to save some
time to present the new material I'm preparing. Since I may have to
present those 8(ish) pages of notes rather quickly, you might find it
easier to follow the discussion if you read over the section on
cross-validation in the article by Efron and Gong.
- July 12, 2002 (4:01 AM)
-
I forgot to give you the solution to Problem 18. Since only 7 students
got it right, you may want to check out the answer I posted on the
homework solutions web page
before doing Problem 19 (since one needs to use the exponential
reference rule to do part of Problem 19).
(I have a hard copy version
of the solution that I can distribute on Tuesday.)
I posted the solution for
Problem 15 as well.
Also, I posted
an explanation of the answers to
Thurday night's quiz, and I
put one more correction to Section I of the class notes on
the corrections web page.
- July 11, 2002 (4:27 AM)
-
I edited
an explanation of the answers to
Quiz #11 to state that I'm not going to count off for the poorly
worded question if you didn't supply the answer I was looking for.
Also, I put one correction to Section I of the class notes on
the corrections web page,
and I added a suggestion for Problem 19 to
the homework web page.
Finally, I added some comments about what I'll cover in lecture today to
the bottom of
the reading guide web page.
- July 10, 2002 (4:55 AM)
-
I posted
an explanation of the answers to
Quiz #11, and
an explanation of the answers to
Quiz #14. Overall, the scores weren't real good, but that's pretty
much been the case all along. I suspect that people do spend a decent
amount of time preparing for the quizzes, and so it must be that they
are a bit trickier than I want them to be. In the end, except for one
guy who may have a perfect quiz average once I drop 4 low scores, most
of you will have similar scores, and so really the quizzes won't
separate you in the determination of the final course grades. I'm still
planning on giving more grades of A and A- than I usually
do, since I think your effort has been good and I consider you to be an
above average group of students. Plus, I want my grading to be fair for
students who came into the course with only minimal preparation (while
satisfying the prerequisite), and that being the case, it's not
surprising that some of you with a lot of courses under your belts will
do well gradewise. One reason for the flood of HW problems distributed
near the end of the course is to give some students a chance to work
hard and earn a good grade despite having made several silly mistakes
previously on the homework. Of course, I hope that some of you who have
done very well on the homework so far will also try the remaining HW
problems, but I won't hold it against you if you choose to coast a bit
over the next week or two. As long as you get close to 160 points on
the HW problems that count towards your grade (after I drop the low
scores like I have indicated), that'll be fine. (Since the HW scores
have generally been good (median score on the problem you turned in Tuesday
night is 14.9 out of 15), unlike the quizzes, I don't think the homework
has been too hard or too tricky, and so I do think it's reasonable
to expect you to be close to 160 points after I drop a lot of low
scores.) It's turned out that the homework problems have been longer
and more numerous than I guessed would be the case six weeks ago, but by
having you work hard on the homework, I can overlook less than
stellar performance on the quizzes and the exam and still give a lot of
good grades --- plus, if you do the minimum number of problems well, I
do think the amount of homework is less than usual for one of my
graduate classes.
- July 9, 2002 (3:29 AM)
-
I updated
information about quiz component of
course grade and
information about homework component of
course grade, now that I have all of the remaining quizzes ready to
go, and I have the last of the HW problems to distribute in class today.
You can find the remining HW problem due dates and new links to hints and
comments about several problems on the
homework web page.
I have also added some more solutions to the
homework solutions web page.
Finally, let me ask that you help me remember to do a few things in the
days ahead: on the days that you turn in Problems 18 and 20, I want to
give you solutions at the end of class (so that they can possibly be of
help to you as you do other HW problems that are related); and help me
remember to do the course evaluation either this Thursday or next
Tuesday (since in the summer the end of the semester seems to swoop in
quickly, and I'm worried that with so many different things in my bags,
the evaluation form packet will get pushed aside as I race to work in
all of the quizzes and the overhead presentations).
- July 8, 2002 (12:14 AM)
- I've added some more
comments about
Problem 14. (The new stuff is at the bottom of that web page, in a
different color from the original comments I posted on June 29.)
(Note: I plan to be at GMU another 24 hours straight, so I may
indeed make more adjustments to the web site later today.)
- July 5, 2002
- I added information about Quiz #14, which I may give as a 2nd quiz
on Tuesday, July 9, if there is time, to the
quiz web page. (I also added a
link to an explanation of a previous quiz to the quiz web page.)
- July 4, 2002
- I posted
HW Problem 18, which is a
simple one that is due on Thursday, July 11. (I will give you the
solution to this problem at the end of class on July 11 if you ask me
for it. You may want to check your answer against the solution before
doing part of Problem 19.) I also posted
some information about Quiz
#11, which will be given at the start of class (so don't be late) on
Tuesday, July 9. I made a slight change to
some information about Quiz
#13, since it is doing Problem 16, instead of Problem 14, that will
help you prepare for the quiz. (For Problem 14, I've worked out the
Satterthwaite approximation in the class notes, but for Problem 16 you'll
need to work it out for yourself, and by doing so, use some knowledge
that is relevant to one of the problems on Quiz #13.) Finally, I updated
the corrections web page.
- July 3, 2002
- I posted
explanations for the correct answers to last night's
quiz.
- July 1, 2002
- I created a
web page pertaining to the
quizzes. Of particular interest to you may be the links to studying
suggestions for the 10th, 12th, and 13th quizzes. (I have yet to decide
what Quiz #11 will cover.)
- June 29, 2002
- I've made some
hints and/or comments about
Problem 14.
- June 27, 2002
- A short while ago a student in the class brought my attention to a
mistake I had in one of my macros --- when I converted my macro that did
the Welch F test for the I = 3 case to handle the I = 6 case,
I forgot to change the 1st df from 2 to
5, and thus my p-value for that test on p. E12f of the
class notes is incorrect.
The correct value should be 0.0567388. (One of the reasons I
didn't give you a bunch of macros like I do in STAT 554 is that I wanted
to have you help me confirm that I got my marcos prepared without bugs.
Of course, I also think it is good practice for you to write some macros
for the various test procedures --- you should learn a lot more by
writing your own macros than by just using ones that you're given.)
I'll note this correction on my
corrections web page, but
since I'm giving you the correction above, there is really no point in
going there now (since I didn't make any other changes today).
- June 26, 2002
- I've made some
hints and/or comments about
Problem 11
and some hints and/or comments about
Problem 12. I'll add similar web pages for some of the other HW
problems later.
- June 25, 2002
- I've updated the
HW web page to show due dates for
several problems that I'll be distributing in class tonight.
Please read the comments I've inserted at the
bottom of the
HW web page. I'll
organize the comments better, and add some hints, a bit later, but if you
read them over in their present form they might prevent some of you from
making some mistakes.
- June 24, 2002
- I've came up with a plan for this week's quizzes.
- Tuesday's quiz will focus on the meaning of a few simple
terms: robust for validity, conservative (test), anticonservative,
homoscedasticity, heteroscedasticity, stochastically larger, and
stochastically smaller. (If you don't know what these terms mean, then
you haven't followed big parts of the lectures very well.)
- Thursday's quiz will focus on some of the distribution
theory and probability underlying parts of Section E of the class notes.
There will be one question where I ask for an expected value, another
where I request a variance, one where I want to know what some function
of random variables converges to (possibly a constant, or possibly a
random variable), and one where I want to know exactly what distribution
some function of normal random variables has. I suggest that you try to
follow the probability given on pages E1, E2, E3, E7, E8, E21, and E22
of the course notes. For some parts, you can take the facts as being
givens (you don't have to know how to derive them), but you
should be able to understand how I go from one line to the next on those
pages.
- June 23, 2002
- I gave some indication of what I'll cover Tuesday (after I finish
through p. E29 of the notes I distributed) in the
reading guide (go to the
bottom to see the plan for Tuesday). Also, I noted some small mistake
in Miller's book on pages 108 and 109. So if you're keeping track of the
mistakes I have found in that book, go to my
web page with comments about Miller's book to see
the mistakes I recently found on pp. 108-109.
- June 22, 2002
- I corrected the macro that I used to produce p. E29 of the class
notes, incorporating a continuity correction for the J-T test.
The new p-value is about 7.2 * 10-5,
instead of 6.3 * 10-5. I'll give you a new (corrected) p.
E29 next class.
- June 21, 2002
- I updated the
corrections to the class notes
web page, posted a solution for Problem 8 on the
HW solutions web page,
posted the due date for Problem 13 on the
HW web page,
and created a
web page which explains the answers to the questions on Quiz #7.
- June 20, 2002
- I'll remind you that a high school graduation is scheduled to start
at the Patriot Center at 7:30 PM, and so you might run into heavier than
usual traffic coming onto campus. Also, in case I forget to tell you in
class, I'm not absolutely certain that I don't have any mistakes in my
Minitab output in the class notes (although I do check things rather
carefully, so I'll guess there aren't many mistakes). I'm having to
work long hours while I'm tired just to get the materials ready for
class each time, so it could be that a mistake is occassionally made.
(For example, I forgot to add the continuity correction when I got the
p-value on p. E29 for the JT test.) So if you check your work
carefully, and you can't match my results, let me know as soon as
possible. Also, check the web site from time to time to see if I
announce any more mistakes.
- June 19, 2002
- 12 of 14 students got HW Problem 7 correct. One of the wrong
answers is close, so I suspect it's off due to sloppiness, but one
student's answer is way off. I've posted the answer, and show how one
can easily get the jackknife estimate of standard error using Minitab,
on the
HW solutions web page. (Once
the data is in a column, one can actually compute the standard error
estimate using a single let command.) My Minitab scheme can be
easily altered to work for the situation of HW Problem 8, so you can use
it to check your work if you wish to do it some other way.
I'm only going to allow about 5 minutes for the quiz on Thursday, so I
encourage you to review
these simple guidelines for
choosing between jackknifing and bootstrapping. The estimation
settings dealt with on Thursday's quiz may be a bit more complex than
those covered on last Thursday's quiz, but if you understand the
guidelines, it should be no problem to correctly deal with the slightly
more complicated settings. (9 of 14 got Tuesday's quiz 100% correct,
and 4 of the other 5 only missed the one question that I commented on.
I'll explain the correct answers in class on Thursday if you remind me
to do so.)
I posted some more
corrections to the class notes
and also pointed out on
this web page
an error in the Efron and Gong article that an
observant student
found (and I also corrected
a very small (almost of the typo sort) error I had on tht web page).
If you handed in HW Problem 8 ahead of time, I will give them
back to you ungraded along with Problem 7 that I've graded, so please
be sure to resubmit it.
- June 18, 2002
- I posted due dates for the 3 problems I'll distribute in class
tonight on the
homework web page.
- June 17, 2002
- I created a
web page with some comments about the Efron and Gong article that
Tuesday's quiz will be based on, and a
web page which explains the answers to the questions on
Quiz #5 (which you took last Thursday, and is similar to the quiz I'll
give this coming Thursday).
To prepare for Tuesday's quiz, read the first 3 sections of the Efron
and Gong article. Don't worry so much about the mathematics or the parts
of those sections that pertain to material that I didn't mention in
class. Also, although you should study Tables 1 and 2 so you can
compare the performances of the bootstrap, jackknife, and parametric
estimators, for the quiz I won't focus on anything specific from the
tables. (But for the final exam, I may have a problem or two which
deals with the tables, so it may pay off later if you spend time now
trying
to fully understand the tables and the Monte Carlo studies done by the
authors.) By looking at my
web page with some comments about the Efron and Gong article, you
might be able to come up with a good guess about the focus of Tuesday's
quiz (which will have 4 or 5 True/False questions that should be
easy, and so I plan to only allow about 5 minutes for the quiz
(so there won't be a lot of time to read the article during the quiz).
For Thursday's quiz, study
this web page and see the
appropriate parts of the class notes for additional details.
Below are some more announcements.
- There will be many high school graduations held at the Patriot
Center this week and next week, and several of them
may interfere with you getting to
class on time if you don't make appropriate adjustments.
Specifically, this Tuesday
a graduation is scheduled for 7:00 PM,
this Thursday
a graduation is scheduled for 7:30 PM,
and next Tuesday
a graduation is scheduled for 7:30 PM.
- If you handed in HW Problems 7 or 8 ahead of time, I will give them
back to you ungraded along with the problems I've graded, so please
be sure to resubmit them.
- I sent out an e-mail to all students in the class in the wee hours
of Saturday morning, and it appears to have bounced back from one
student's mail service, so let me know if you didn't get it and/or want
to add another e-mail address to my mailing list.
- I made one more correction (to page E6) on the
corrections web page.
- I slightly improved the
web page pertaining to choosing
between jackknife and bootstrap estimates, stating that for a linear
plug-in estimator an estimate of the bias isn't even necessary because
such an estimator is unbiased.
- I added one more book as a reference about bootstrapping to the
reading guide.
- June 15, 2002
- I updated the
reading guide to indicate what
I'll cover this coming week (go to the bottom of the web page to see
where I indicated that I'll cover certain parts of Ch. 3 of Miller's
book) and also include some more references on bootstrapping,
and I updated the
web page listing corrections to the
class notes (with all of the new corrections being at the bottom of
the web page). (Please let me know if you spot any other errors in the
notes, or other materials that I give you.) I also fixed a couple of typos
(nothing of importance) on web pages linked from
the syllabus, as well as corrected the time
period for the final exam on a
web page linked from the syllabus (since
the exam period goes until 10:15 PM).
I advise that you review the material from Sec. C of the class notes
about choosing between the jackknife and bootstrap estimates of bias and
variance (and standard error). (Go
here to see a short summary.)
No one got all 5 parts of Thursday's quiz
correct. As I said in class, next Thursday (after you've gotten back
the quiz you took) I'll give a similar quiz about choosing between the
jackknife and the bootstrap. Please remind me on Tuesday to explain the
reasons for the quiz answers after I give back your papers to you.
- June 11, 2002
- I updated the
homework web page
and the
reading guide,
and added a
web page listing corrections to the
class notes
and a
homework solutions web page.
- June 10, 2002
- I only plan to add one more page to my notes on bootstrapping, and
one more page to my notes on jackknifing, and so we should finish up
those topics (for now --- they will be revisited later) in about an
hour during Tuesday's class. The next topic will be dealing with
dependent observations, the subject matter of Sec. 1.3 of Miller's book.
In the process of discussing dependence due to blocking, I'll get into
subsection 3.5.4, which makes use of material on pp. 95-96. When you
have time, you might want to look over any pertinent comments I have about
Chapters 1 and 3 in
my
notes about the Miller book. (Just
tonight I have pointed out a mistake on p. 104 that I hadn't noticed
before.) Although my lecture material on dependence gets a bit messy in
places, I'm not going to have you do any HW problems based directly on
the messy part. What I hope is that the main message comes through:
if the data comes from depenent random variables, and appropriate
adjustments are not done, then tests and confidence intervals can be
rather misleading.
- June 7, 2002
- I edited the
homework web page to include for due dates for the problems I've
distributed. Also note, in case there is any confusion about what I
announced in class, although I plan to give quizzes next week
like those described
on the sheet containing HW problems 5, 6, 7, and 8, I will reverse the
order, so that the quiz I have described for Tuesday will be given
Thursday, and the one described for Thursday (based on the first 2 pages
of the Science article), will be given on Tuesday.
- June 1, 2002
- The scores on the first quiz weren't so good. The next quiz will
be 10 True/False questions based on these two topics: (1)
the improved estimate of bias; (2) bootstrap confidence intervals
based on approximate normality and using the bootstrap estimate of
standard error. Since people did miss more than I had expected them to
on the first quiz, I'll go over the quiz at the beginning of class, and
give a 5 minute review of the basic bootstrap estimate of bias --- maybe
one more time through the explanation will work wonders for some
people's understanding. Of course, you'll get more out of the review if
you spend some time thinking about the estimate of bias on your own. Go
here to see a short
explanation that may be helpful.
Also, note that I updated the
homework page to show the due
dates for the problems due on Thursday. (Two problems are due Tuesday,
and two are due Thursday.)
- May 29, 2002
- I edited the
syllabus and the linked info about the quizzes to reflect the fact
that part of the 1st quiz will be closed books and notes. (The
rest of the quizzes will be open books and notes.)
- May 28, 2002
- I edited the
syllabus, adding three links that give more
information about the grading. Also, I put the due date of the first HW
problems on the
homework web page.
- May 24, 2002
- I edited the
reading guide to give you a
better
indication of what you can read to help prepare for the first lecture
now that I've changed my mind about the order in which I will cover
topics during the first 2 or 3 weeks.
Really, it's not that important that you read anything prior to the
first lecture, but if you have time, check out the article that I list
which is available in e-journal format.
- May 11, 2002
- I edited the
reading guide to give some
indication of what you can read to help prepare for the first lecture.
- April 26, 2002
- If you want to take the course this summer, be
sure to register for it by May 13 --- if too few students have
registered, it may be cancelled.