Experimental Design Fall 2016
T-Th 3:00 - 415 Wang 2555
Instructor: Professor Tom Kuczek (firstname.lastname@example.org)
Office: 152 HAAS, 765-494-6051
Office hours: Monday 1-2, Thursday 1-2 or by appointment. Web Address: http://www.stat.purdue.edu/~kuczek/stat514
Teaching Assistant: Lin-Yang Cheng (email@example.com)
Office: MATH 537
Office hours: Tuesday 1:45-2:45, Wednesday 1:45-2:45
· Text: Design and Analysis of Experiments by Douglas Montgomery (5th, 6th, 7th or 8th edition ok). Copies are on reserve in the Math Library on the 3rd floor of the Math Building. The Library no longer has free access to the text.
Experimental Design is a widely used tool for research in Industry and Academia. Its origins lay in the applications of Statistics in Agriculture. Current application areas also include the Chemical and Pharmaceutical Industry, Manufacturing in general, the design and analysis of Clinical Trials, Quality Control, etc.
General Course Policies:
1. Attendance and Excused absences: While attendance will not be recorded due to the nature of the distance class, material covered in class will be on the exams. If an event occurs which requires an absence (for on campus students) that affects turning in an assignment or taking an exam, please contact me and we will make appropriate arrangements.
2. Answering questions via email is not a problem, on weekends I unplug but will check email for questions. Please Put STAT 514 in the subject header since these get priority.
3. Students arriving late/leaving early. Sometimes this cannot be avoided due to scheduling, but please be considerate of others.
4. Technical issues: if there are technical issues, for example, using software from ITaP remotely, please bring these to my attention as soon as possible so any issues can be resolved quickly.
5. Class participation/preparedness: Though the class is online, there is some level of interaction within the classroom for those who are on campus. Don’t let the online nature of the class keep you from interacting.
Academic dishonesty: Purdue policy on academic dishonesty.docx
Your grade will be determined by your total out of 300 possible.
Exam I: Thursday of week 8
Exam II: Thursday of week 15 (week before the last week of classes).
Exams are based primarily on lecture material and designated reading material. You are responsible for what is discussed in class. There is a hyperlink to a folder entitled “Old Exams Stat 514” below. The answers are not given (so you have to actually work out the problems), but please feel free to ask me about any of your answers, or if you can’t get the answer, by e-mail or at office hours.
Homework submissions for on campus students:
· Please do not turn in a stack of output to HW questions, you should copy and paste the relevant parts of the output into a write-up of your HW.
· Please turn in a Hardcopy of your HW in class. It is simply a question of document control (that’s what we call it in Quality Control).
· If you do not turn HW in class turn it in to the Grader (a copy to the grader’s mailbox is fine, a pdf by email also ok), Never to me. Late HW is ½ off.
· Note: while individual homework problems will likely be graded out of 10 points, at the end of the semester they will be normalized to a 90 point total!
Homework submissions for off campus students:
Off campus students will submit their HW using Blackboard. The instructions are in this link: Submitting an assignment in Blackboard.docx
Statistical Software and Software Resources.docx Online help from the Statistical Consulting Service (SCS)
An on campus SAS evening help session is open to all students taking courses involving SAS coding.
SAS help sessions for Fall Semester 2016:
Date Time Location
Wed 08/24 - 11/16, 2016 6:30p - 8:30p BRNG B286
Wed 11/30 - 12/07, 2016 6:30p - 8:30p BRNG B286
Who to ask about what:
1. Software questions can be directed to the Statistical Consulting Service (SCS) at firstname.lastname@example.org
2. Questions about HW or HW grading should go to the TA at email@example.com . Software questions related to the HW can also go to the TA
3. Questions about course content, exams and exam grading go to me firstname.lastname@example.org
Weekly Schedule for reading and homework:
Week 1: Aug. 22 - 26.
· Read Chapter 1 (not too exciting background on experimentation).
· Read Chapter 2, relevant sections. Concepts of the Statistical model, hypotheses to be tested, conclusions, Type I and Type II error, power of a Statistical test and sample size are key concepts for two sample and paired t-test.
Week 2: Aug. 29 - Sept. 2
HW 1 due Thursday Sept. 1: Problems 2.27 and 2.31 (do only parts a) and b) for 2.31).
problem 2.31.xls this format is easier to input into SAS
paired data.xls (for 2.27)
· Begin One-way ANOVA
· STAT514-notes-0901.pdf Notes from Thursday, still need to work on my writing :-/
Week 3: Sept. 5 - 9.
· Monday is Labor Day.
· Read Chapter 3, relevant sections.
· Important Concepts: the model assumed, assumptions, expected mean square for fixed model, post F-test comparison of means, experimentwise vs. comparisonwise error rates, normality plots, contrasts..
Week 4: Sept. 12 - 16.
HW 2 due Thursday, Sept 15: Problems 3.5 (for part c, you can use any comparisonwise procedure), 3.6 a&b.
HW2 problems.pdf For part b), a Box plot will suffice. For part c), any Comparison-wise procedure will suffice. In SAS, in the Means statement after the ANOVA write /LSD or /Tukey to get the appropriate means comparisons.
· Example Problem 3_24.pptx These are the annotated slides from Tuesday.
· Finish One-way ANOVA, contrasts, etc.
· ANCOVA_annotated.pptx comments during class
· Read Chapter 5, Sections 5.1, 5.2, 5.3 and 5.4.
Week 5: Sept. 19 - 23.
HW 3 due Thursday, Sept. 22: Problems 5.6 and 5.10. For 5.10b, just estimate cell means.
Week 6: Sept. 26 - 30.
HW4 due Thursday Sept 29: For problem 5.10, now assume that the factor Glass type is Random and Temperature is fixed.
Use the algorithm to compute the Expected Mean Squares.
Redo the analyses with the “correct F-tests” and range tests on Temperature (if necessary).
· ANOVA and EMS.docx Things related to analysis of Mixed Models Dr. Tom didn’t tell you…..
Week 7: Oct. 3 - 7.
HW5 due Tuesday October 4:
Problem 5.19, (assuming operators random). Make sure that you calculate the EMS with the algorithm for the correct F-tests. Indicate if any post F-test analysis is necessary.
HW6 due Tuesday October 4:
Problem 5.21, Temperature and Pressure are fixed effects. Part I, assume days fixed, but delete the highest order interaction from the model so that it ends up as the error term. Part II, assume days random, but keep all terms in the model and do the correct F-tests. Make sure you calculate the EMS with the algorithm. Indicate if any post F-test analyses if necessary.
· Thursday: Review for Exam I
Week 8: Oct. 10-14.
October 10-11: OCTOBER BREAK
October 13, Exam I
· TypeI and TypeIII Sums of Squares.pdf The Dummy Variable Model in SAS (and other software packages)
Week 9: Oct. 17-21.
HW 7 and HW 8: due Thursday
· Problems 4.3 and 4.33.pdf Make sure that you write out the Model, Hypotheses tested, and conclusions for each problem. You don’t need to work out the EMS for these two problems.
· BIBD and Adjusted Sums of Squares2.pptx annotated from lecture.
· problem 4.3.xls :
· LS Means.pptx Means vs. LS Means
· Repeated Measures.pptx Hicks Example
Week 10: Oct. 24-28.
Repeated Measures and Nested Factorials.
HW 9 due Thursday, November 3:
On Dr. Tom’s Imaginary Farm, Dr. Tom raises Unicorns (he hasn’t gone Vegan yet) and wants to know the best way to cook the meat. He is interested in the factors Tenderizer and Roasting Time. From six Unicorn carcasses he selects a large piece of meat which he divides into three samples. Each sample receives one of three Tenderizer treatments. After the Tenderizer treatment is completed, the samples are divided into four subsamples, each of which is cooked for one of four Roasting Times. Afterward, the tenderness of the cooked Unicorn meat is tested for tenderness which is the Response Variable, Y.
1. Draw the Layout, including Unicorns (the Subjects).
2. Write out the Model.
3. Write out the ANOVA table with Source and df. Make sure that the Between Subjects terms are first and the Within Subjects terms are next.
4. Work out the EMS with the algorithm.
5. Do the analysis with the correct F-tests, stating the hypotheses, F-statistics and conclusions.
6. Do any necessary Post-F-test analysis of Means for the fixed effects.
8. If Unicorn had been left out of the Model, how would this have changed the analysis?
Read Split plot material in book, Chapter 14.
Read first six pages of: Split plot example.pdf
Week 11: Oct. 24-28.
Homework problem, due Thursday, November 10: For the data of the Box split plot experiment:
Including the whole plot in the model:
1. Write out the Model.
2. Work out the EMS with the algorithm.
3. Do the analysis with the correct F-tests, stating the hypotheses, F-statistics and conclusions.
4. Plot the interaction term means.
5. Redo the analysis above, except leave the Whole Plot term out of the model.
6. Explain why the incorrect analysis gives the wrong conclusions.
Week 12: Nov 7-11.
Finish Split Plot material.
· Read Latin Square and Graeco–Latin Square material in the book.
· HW 10 (two problems) due Thursday Nov. 17
· Crossover Designs
· Review for Exam II.
Week 13: Nov. 14-18.
Read Chapter 6.
Week 14: Nov. 21-25. Wed-Sunday Thanksgiving Vacation.
· Exam II answers.pdf except EMS. Curve discussed in class.
Week 15: Nov. 28-Dec. 2.
Review for exam in Tuesday class.
Exam II Thursday Dec. 1
2k designs and analysis. Ch. 6 in textbook.
Problems 6.17 and 6.20 due Thursday, December 8, as follows. For the data of 6.17, construct a normal plot and choose a model based upon a plot of the effects. For the data of 6.20, analyze the data using only the main effects and two way interactions in the model.
Read Chapter 7, sections 7.3 and 7.4. Read Chapter 8, sections 8.2 and 8.3.
Week 16: Dec. 5- 9 Response surfaces intro
Finals Week: Dec.14-19.
Old Exams Stat 514 Includes older exam II’s.
Only calculators without communication capabilities will be allowed. Any formulas which are needed will be provided.
1. Introduction to the Study of Experimental Medicine by Claude Bernard (Introduction à l’étude de la medicine expérimentale de Claude Bernard circa 1860). An absolute classic but outstanding reading for anyone interested in the (serious) application of the scientific method to Medicine or Biology in general.
2. Placebo Effects: Understanding the mechanisms in health and disease by Fabrizio Benedetti (2008). So you think you understand the placebo effect huh? Hah, read this!
3. Response Surface Methodology by Ray Myers and Doug Montgomery. Probably the best book on the subject I have seen.