Announcements

New! Grades have been submitted to the registrar (7:30 pm, Wednesday, May 20). According to the computer, they should be available on-line within 24 hours (by 7:30 pm, Thursday, May 21).

New! Grade distributions for final exam posted.

New! Final exams are graded and can be picked up from the instructor 356 Ford Hall.

New! Solutions to the final exam posted.

The instructions for the exam say You may use three 8 ½ × 11 sheets of paper with formulas, etc. You may also use the handouts on brand name distributions and Greek letters.

The final exam has 8 questions; 4 are new material like problems on homeworks 9 thru 12; 2 are like questions on the 1st midterm; 2 are like questions on the 2nd midterm.

The final exam will be at the scheduled date and time (10:30am–12:30pm Saturday, May 16) but the room has been moved because Hasselmo Hall is locked on Saturday. The exam will be in Moos Tower, Room 2-690. It is on the concourse just below street level.

Homework 12 solutions posted. That all of them.

Errors found in class Wednesday but forgotten until Friday corrected: slide 12 (p = 6 changed to p = 7 to match the web page and correspondingly five predictors to six predictors), slide 14 (BIC did not select the correct model), slide 18 (p = 5 changed to p = 6 to match the web page and correspondingly four predictors to five predictors), and slide 20 (p = 8 changed to p = 10 to match the web page and correspondingly seven predictors to nine predictors).

Office hours for Geyer next week will be 2:30–3:20 MWF. No double hours on Wednesday. Office hours for Geyer Friday, May 8 will be 2:30–3:20 and 3:35–4:25. The last homework will be due by 4:30 pm, as usual.

Homework 9 solutions and Homework 10 solutions and Homework 11 solutions posted.

Problem 12-5 modified to be like the computer examples about model selection. Former part (b) is canceled. Also use the R function regsubsets rather than leaps. (regsubsets is newer and better.)

Homework Assignment 12 posted. Due last day of class, Friday May 8.

The version of deck 6 used for the last part of class is now on the web, with the error on slide 34 noted in class fixed (the way powers of n appear in the third displayed equation).

Errors and omissions found in class on slide 4 (if A M is full rank added), on slide 5 (B changed to BT in three places), and on slide 20 (n changed to 1 in three places) fixed.

A student noticed that Rweb complains about the dataset

http://www.stat.umn.edu/geyer/5102/data/prob11-4.txt

complaining

Rweb:> attach(X) 
 
	The following object(s) are masked from package:datasets : 
 
	 pressure  

if this dataset is even loaded. Don't worry about the warning. It just says that some dataset named pressure is unavailable while these data are loaded. That's no problem, you don't even know what that other dataset is.

Error found in class on Homework Problem 11-9 fixed (needed hat on theta in part (b)).

Errors noted in class on slides 60–61 (proof of lemma proved the wrong two cases) and on slide 63 (statement of theorem omitted σ2 in denominator of first two statistics) fixed.

Error noted in class in section on F tests of model comparison on linear models web page fixed: β3 = β4 = β5 = 0 changed to β4 = β5 = β6 = 0.

Errors on Homework 1 noted after class by students. In Problem 10-1, the data set is described wrong. The variables are x1 through x5 and y. There is no x6. Fixed. In Problem 10-2, the same mistake is also fixed. In Problem 10-3, the same mistake is also fixed. Moreover In Problem 10-3, part (a), another mistake (given x1 and x1 should be given x1 and x2) is also fixed. In Problem 10-4, introduction, part (a), and part (b), the same two mistakes as in problem 1-3 are fixed. In Problem 10-6 the same mistakes as in problem 1-1 (there is no x6 and there are five predictor variables not six) fixed.

Errors on slides noted in class on slide 54 (two subscripts new weren't subscripts) and after class on slide 56 (title should be Prediction Intervals rather than Confidence Intervals) fixed.

Error on web page noted in class on section on tests about correlation coefficients1 should have been β2) fixed.

Error reported by a student after class on deck 5, slide 10 fixed (E changed to var).

Solutions for second midterm posted.

Grade distributions for second midterm posted.

In homework problem 9-4 it was intended that only the success probability be considered an unknown parameter. Now fixed.

Homework Assignment 9 posted. The due date is Wednesday, Apr 15, 2009. There is one error on the paper copies handed out in class that has been corrected on the PDF linked above. In problems 9-8 and 9-9 the prior distribution is denoted Γ(3, 3) on the paper copies and Gam(3, 3) on the PDF, just like everywhere else in this course.

The TA found an error on the solution for homework problem 8-9. The exponent − 2 α − 5 was incorrect. It should be − 2 α − 1. Now fixed.

Error found after class on Two Sample, Two Tailed Bayes Test computer example fixed. The same error is on slide 108, deck 3, also fixed. The error is subtle. The distribution of the data given model m1 and the parameter p is

C(n1, x1) C(n2, x2) px1 + x2 (1 − p) n1 + n2x1x2
not what was there before
C(n1 + n2, x1 + x2) px1 + x2 (1 − p) n1 + n2x1x2

Some Bayesians in my department expressed no surprise at the fact that the Bayesians are much more conservative about tests than frequentists. You've just rediscovered Lindley's paradox one of them said.

A student found an error on the solution for homework problem 8-3. The term λn α was omitted from the likelihood and when it is put in it changes the answer to Gamma(n α + &alpha0, n xn + λ0).

No homework this week, have midterm instead (Wednesday, April 8). Consequently, the office hours for Geyer on Wednesday are moved to Tuesday April 7. Two office hours 2:30–3:20 and 3:35–4:25. This change is for this week only. Back to regularly scheduled office hours next week.

A student found an error on deck 2, slide 214. Now fixed. Now says X(k + 1) < θ0 < X(nk). Before k + 1 was k − 1.

Error found in class on One Sample, One Tailed Bayes Test computer example fixed.

The infelicities found in class on slides 99–101 fixed: p(m) changed to p(m | x).

The midterm exam Wednesday, April 8, 2009 is not cumulative. It will cover everything since the first midterm: homework assignments 5–8 and slides starting with slide 147, deck 2 and going through slide 47, deck 4.

Solutions for homework assignment 8 posted.

The errors found in class on slides 84–85 fixed (alphas should be +1⁄2 not −1⁄2).

Solutions for homework assignment 7 posted.

Because I covered the wrong material in class Monday, we didn't cover the material needed to do problems 8-11 and 8-12. They are canceled for this week. If you have already done them, they will be on homework 9.

The copy of Homework Assignment 8 that has been on the web page has been incorrect until now. The copies that were handed out in class were correct, but the correction did not get on the web. Now corrected. In particular, the data distribution in problems 8-5 and 8-7 is Geo(p).

The errors found in class and after class by a student on slide 49 (minus sign that shouldn't have been there) and on slide 51 (absolute value signs missing in two places) fixed.

The error found in class on slide 112 fixed (square that shouldn't have been there).

Homework Assignment 8 posted. The due date is Wednesday, Apr 1, 2009.

The errors found in class on slide 99 fixed (convex changed to concave).

While doing the solutions for homework assignment 6, it was decided that questions 6-1, 6-2, 6-3, 6-4, and 6-6 would read better if they said X1, X2, …, Xn instead of X1, X2, …. This is the same issue that caused trouble in 6-5. Hope everyone did these problems for sample size n anyway.

Solutions to Homework Assignment 6 posted.

The error found in class on slide 92 fixed.

Solutions to Homework Assignment 5 posted.

Web page on sign tests modified. It had two headings lower-tailed tests, one of which should have been upper-tailed tests. Now fixed.

While doing solutions for homework 5 it was noted that problem 5-5 was worded incorrectly. Fixed. (Too late, of course, hope everybody figured this out.)

Hint: for problem 6 13. See slide 23, deck 1.

The errors and omissions found in class on slide 70 and slide 75 fixed.

Error on homework assignment 7. The suffix for URL datasets is .txt but the dead tree format assignment again wrongly has .dat Please fix.

Error on the brand name distributions handout, Student's t Distribution in the first bullet under Relation to Other Distributions, the first letter should be X rather than Z.

Some of the students pointed out an error (or at least ambiguity) with homework problem 6-5. Both sample sizes are supposed to be the same. Now fixed.

The errors and omissions found in class and by students on slide 17 — where s < x have the inequality reversed — and on slide 23 — where nu hat should have a subscript n in three places — and on slide 28 — where nence should be hence — and on slide 35 — where then should be than — are now corrected.

Also there are two office hours for Geyer today March 6: 2:30–3:20 and 3:35–4:25.

xkcd does statistics

There is an error on the paper copies of Homework Assignment 6 handed out in class: Problem 6-7(c) should say for problem 6-3, considering μ the unknown parameter and σ known. (We don't know how to do two-parameter problems yet.) This has been fixed on the PDF.

The errors and omissions found in class on slide 16 — where convexity should be concavity — and on slide 17 — where f'(x) ≥ f'(s) = 0 should be 0 = f'(x) ≥ f'(s) — and on slide 11 — where not containing ν should be not containing μ — are now corrected.

Those who were not in my section last semester may not know that (1) the final is cumulative and so will cover some of the material on the first midterm. My policy is to ask questions similar to the ones the class had the most trouble with. For this exam, question 4 was clearly the hardest, and questions 3 and 5 were about tied for next hardest. So those would be the ones to especially make sure you understand.

Grade distribution for the first midterm posted.

A student found multiple errors on slide 228, deck 2 where every little n should be a big N = n (n + 1) ⁄ 2. Now fixed.

Solutions to the first midterm posted.

Seen on reddit: an article in PLOS Medicine Why Most Published Research Findings Are False.

A student noticed an error in the slide 66, deck 1. Where it said j − 1 in two places under a summation sign, it should have said j = 1. Now fixed.

A student noticed an error in the homework solutions 1, problem 1-6. The terms (1 − v)ν1 − 1 and (1 − v)ν1 − 2 that occur in two integrals should have been (1 − v)ν2 − 1 and (1 − v)ν2 − 2, that is ν1 should be replaced by ν2. Now fixed.

The rules for the midterm say you can use the brand name distributions handout plus one 8 1/2 by 11 sheet of paper with formulas or whatever on it. You can also use a calculator. Nothing else.

A student noticed an error in the homework solutions 2, problem 2-6. I forgot a square root in calculating σ from σ2. What was ν ⁄ (ν − 2) should have been √ν ⁄ (ν − 2) and is now fixed to be so.

The midterm exam Wednesday, February 25, 2009 will cover everything covered by homework assignments 1–4, the solutions for which have been posted. In the slides, this is everything in deck 1 and everything in deck 2 up to but not including hypothesis tests and power, so the last slide covered is slide 146, deck 2. Hypothesis tests and power will be on the second midterm.

A comment about mathematical symbols in R plot labels has been added to the computer examples for hypothesis tests and power web page.

The error found in class on slide 193 where zα should have been zα ⁄ 2 has been corrected.

New computer example page on hypothesis tests and power.

Homework assignment 4 changed too late to matter. I didn't notice until doing the solutions that two of the URL's had the suffix dat when the files had txt.

Homework solutions for Homework Assignment 4 posted. Username and password required to see the solutions, unchanged since last semester.

New computer example page on confidence intervals.

The errors and omissions found in class on slides 150–151 about which of Xi and Yj was treatment and which control and which way the inequalities go in specification of the hypotheses are now corrected.

Homework solutions for Homework Assignment 3 posted. Username and password required to see the solutions, unchanged since last semester.

The errors and omissions found in class on slide 133 and on slide 135 have been corrected. Several spelling errors were also corrected.

Homework solutions for Homework Assignment 1 and for Homework Assignment 2 posted. Also, the Homework Solutions link in the navigation to the left now works. All homework solutions will be posted there. Username and password required to see the solutions, unchanged since last semester.

The errors and omissions found by a student on on slide 64, deck 1 and on slide 74, deck 1 have been corrected. The first problem was (XiXn)2 that was missing the square. The second problem was Zn that was missing the n.

The errors and omissions found in class on slide 105 and on slides 108–110 have been corrected.

The errors and omissions found in class on slides 84–85 and on slide 92 have been corrected.

For homework problem 3-1 it helps to know the definition of scale family. See 5101 deck 3 slides 143–145. A scale family is the special case where the location parameter is fixed.

The errors and omissions found in class on slide 37 and on slide 39 have been corrected.

The error found in class in the proof for the asymptotic distribution of sample quantiles is now fixed. The purported proof only took one slide (was 115), but the careful proof now takes three deck 1, slides 115–117 so two slides have been added to deck 1 at that point.

The error found in class on slide 33, deck 2 is now fixed.

The errors found in class on slide 101, deck 1 was forgotten but discovered again in the review session. It is now fixed.

The Exam Dates link in the navigation to the left now works. The dates for the midterm exams are now scheduled.

I forgot to say Wednesday that the old class notes people signed up for were done and in the copy center near The Cube in Coffman union.

Homework Problem 1-3 was incorrect and has now been fixed. The MAD is the empirical median of the values |xi - m| where m is the empirical median. The absolute value sign here was omitted.

Homework Problem 1-2 was incorrect and has now been fixed. The n on the right-hand side of both the empirical mean and empirical variance didn't belong there and has been removed.

The errors and omissions found in class on slide 50 and on slide 60 have been corrected A student reported another error after class, which has also been fixed. The squares were missing in

E{(Xa)2}
and
En{(Xa)2}

on slide 6 and on slide 7 and on slide 62. All have now been fixed.

Anyone who is curious about the Gram-Schmidt orthogonalization process used but not defined on slide 70 of deck~1 of this year's notes can look at some old notes which were once handed out in 5102 that explain it. You do not need to know any of this for homework or exams in this course.

Those who were not in this class last semester should look at the material on existence of moments from last semester, which are 5101 Deck 4, Slides 2–32.

For anyone who did not get a brand name distributions or greek letters handout, they are linked under the Class Notes link in the navigation.

Those who were not in this class last semester don't know that the username and password to look at exams and homework solutions from last semester. They will be given out in class Monday. They are the same as last semester. Everyone who was in this class last semester knows them.

Those who were not in this class last semester don't know that the scribbles made in class also go on the web pages under the Course Slides link in the navigation.

The answers to the questions about photocopied lecture notes is that they could be bound (about $2 extra) or three-hole punched. Preferably one or the other, not some of both. Think about whether you want them — the old notes under the Class Notes link in the navigation — because we will do the sign-up for the copy center Monday.

Homework 1 will be due in class 9:55 am Friday, January 30. You cannot hand it in later than 9:55 (the usual policy is until 4:30 but will not be in effect for this first assignment) unless you have made a special arrangement with Mr. Zhou.

There will be class Friday, January 30, and the instructor will be the TA, Mr. Zhou.

The instructor (Geyer) will be out of town from the afternoon of Wednesday, January 28, 2009 through the rest of the week. Consequently, his office hours Wednesday, January 28 and Friday, January 30 are cancelled. The current plan is that there will be a substitute lecturer for the class Friday, January 30. More on this later.

Spring semester begins on a Tuesday. The lab sections (2 and 3) for our class scheduled to meet that Tuesday (Jan. 20) will not meet. The first class will be the Wednesday (Jan. 21) lecture section.

The General Info link in the navigation to the left now works. Everything is the same as for 5101 fall semester, except as noted below.

The Course Notes link in the navigation to the left now works. Everything is the same as for 5101 fall semester, except the very incomplete notes started but not continued after the first week have been deleted.

The Course Slides link in the navigation to the left now works.

The Computer Examples link in the navigation to the left now works.

We will be using the old notes as a partial textbook. There will be a sign-up sheet the first day of class for those who want a copy photocopied by copying services. It will be 3 cents a page (about $15). We will cover more or less all of that material starting with Chapter 7. We will also cover some other material which will be on the slides only.

Exam Dates

Exam Date and Time
First Midterm Wednesday, February 25
Second Midterm Wednesday, April 8
Final Exam 10:30am–12:30pm Saturday, May 16

The final exam time is set by the U.