PSY 626: Bayesian Statistics for Psychological Science

Fall 2020
Days/times: Tuesday, Thursday / 10:30 am - 11:45 am
Location: Online

Update

August 25: I changed the link to the WebEx meeting, below. This might work better.
August 25: A link to the recording of today's lecture is now available, below.
August 27: The TA set up some instructions for installing software we will (eventually) use on a Scholar cluster. It also includes some tips for other systems. If you have problems, please contact mkon@purdue.edu.
September 1: ITP may still be setting up student access to the Scholar cluster. If you do not get it to work by Thursday, let me know and I will reach out to them.
September 9: Homework 1 is available, below.
September 22: Despite the problems we had this morning with WebEx, we will try to use it again on Thursday. Skype will be our backup plan.
October 12: I forgot to put a time for submitting homework 2. Send it to Maria by midnight on Friday, October 16.
October 28: Each student will prepare and present a "final project" for PSY 626. Using a data set, you will run a Bayesian analysis (exactly how is up to you). Ideally, you have your own data set to use for this investigation, but if you need a data set, I can provide one for you. Although the end of the semester is several weeks away, it seems prudent to start scheduling the student presentations. I plan to have two presentations each class period. Please sign up for a day at a Google doc by putting your name next to the date you are willing to present in the Speaker 1 or Speaker 2 column.
November 5: In response to a question after lecture, I mentioned a statistics focused blog hosted by Andrew Gelman. It is Statistical Modeling, Causal Inference, and Social Science. It discusses a variety of statistical-related topics. Gelman is a prominent Bayesian. Many of the discussions are quite interesting.

Instructor:

Name Office Email Phone Office hours

Greg Francis PSYCH 3186 gfrancis@purdue.edu 494-6934 MWF 2:30 - 3:30 pm

Office hours: Virtual office hours will be held 2:30-3:30 pm (US Eastern time) via WebEx. If WebEx asks for a meeting ID, use: gfrancis.

Materials (lectures, readings, datasets, code):

Class recordings:

August 25, 2020.
August 27, 2020.
September 1, 2020. (The recording starts 5-10 minutes in the lecture; I just forgot to start recording until then.)
September 3, 2020.
September 8, 2020.
September 10, 2020.
September 15, 2020.
September 17, 2020.
September 22, 2020.
September 24, 2020. (Sorry, the lecture did not get recorded. I thought I had pressed the record button, but apparently, I had not.)
September 29, 2020.
October 1, 2020.
October 6, 2020.
October 8, 2020.
October 13, 2020.
October 15, 2020.
October 20, 2020.
October 22, 2020.
October 27, 2020.
October 29, 2020.
November 3, 2020.
November 5, 2020.
November 10, 2020.
November 12, 2020.

Text:

McElreath, R Statistical Rethinking: A Bayesian Course with Examples in R and Stan. Try to get the second edition. Ordering information and code examples are at the book web site.

In case you do not yet have the textbook, Chapters 1 and 2 of the textbook are on-line.

Lectures:
Lectures will happen online via via WebEx. If WebEx asks for a meeting ID, use: gfrancis. I will record the class meetings and make them available. Try to attend the (online) class if possible, so that we can address questions.

General plan: The course will explain why you might want to use Bayesian methods instead of frequentist methods (such as t-tests, ANOVA, or regression). The general plan is to:

Explain some problems/difficulties with frequentist methods: Publication bias, optional stopping, questionable research practices.
Discuss differences between hypothesis testing and prediction: mean squared error, shrinkage.
Discuss methods for prediction: likelihood, AIC, BIC, cross-validation.
Explain the basic ideas of Bayesian methods: non-informative priors, informative priors.
Provide hands-on examples of applying Bayesian methods: Bayes Factors, hierarchical models.

Throughout, we will be using computer programs to demonstrate the ideas. There will not be any proofs.

Class home page: The home page for this course is http://www.psych.purdue.edu/~gfrancis/Classes/PSY626/indexF20.html From this page you can download lecture notes, view the class schedule, view current grades, and connect to the various homework laboratory assignments.

Homework: Assignments will be due approximately every two weeks. The intention is to use the homework assignments as a way of practicing the concepts we discuss in class. They will be graded, but only to insure that students actively participate.

Homework 1: as PDF, as MS Word, ComputePower.R,
Homework 2: as PDF, as MS Word, SleepySubjects.csv,
Homework 3: as PDF, as MS Word, LookDontType.csv,

Project: In the last two weeks, students will present (document updated 29 October 20202) a Bayesian (or related) analysis of some of their own data. If you do not happen to have a data set, we will get one for you.

Assumed background:

It would be nice, but not necessary, if you had some previous exposure to calculus.
Doing any kind of Bayesian analysis requires some programming. We will be using the free R software. Many people like the R studio program. You do not need to be an expert programmer, but if you have little programming experience, you will have some catching up to do.
Students should have experience with typical statistical methods (t-test, ANOVA, regression).

Teaching Assistant:

Name Email Office hours

Maria Kon mkon@purdue.edu MWF 9:30 - 10:30 am (via Microsoft Teams

Please contact the TA if you cannot meet during office hours to schedule an alternative time.

Name	Office	Email	Phone	Office hours
Greg Francis	PSYCH 3186	gfrancis@purdue.edu	494-6934	MWF 2:30 - 3:30 pm