PSYC272  Seminar in Bayesian Statistics

http://psy2.ucsd.edu/~dhuber/bayesian_stats.html

 

 

                                                                  Class:                    Wed  10-12:50      Room:                        MNDLR 1507

 

Professor:  David Huber                          Office Hours:        Tu 10-12               Office:                        5137 McGill                    

E-mail:  dhuber@ucsd.edu                                                                                  Phone:                          822-1630

 

 

Course Description:

Bayesian Statistics offer an alternative to traditional null hypothesis testing. Traditional significance tests of ANOVA, correlation, and t-tests calculate the probability of observing the results you found (or results more extreme than what you found), under the assumption that there were no effects. But what you really want is the opposite. You don't want to know the probability of your data under the null; instead you want to know the probability that there was or was not an effect. This can be done with Bayesian statistics and, furthermore, it can be done in the framework of our favorite traditional analyses (correlations, t-tests, ANOVAs, etc.). In this course we will cover the mathematics behind Bayesian statistics and we will use the R programming language to learn how to do Bayesian analyses.

 

 

Readings:

Kruschke, J. K. (2011). Doing bayesian data analysis : a tutorial with R and BUGS. Burlington, MA: Academic Press.

http://www.indiana.edu/~kruschke/DoingBayesianDataAnalysis/

Available on Amazon for $74.68. 2 days shipping is $11.98, but free with Amazon Prime.

 

Requirements:

 

   R programming

The textbook is full of source code for implementing Bayesian statistics using R (and WinBUGS). As you read through the chapters each week, keep R open on your computer, and try out some of the analyses. Once you’re done reading the chapters for that week, save the history of your commands, and e-mail the results to me in advance of class. There is no right or wrong way to do this and no guidelines on how much or how little you need to do. I’m just looking for evidence that you tried it out.

 

   Student Led Discussions

Each student will lead the class in covering one chapter. This will take place during one of the two hours of class. You do not need to fully understand the chapter. However, it is your responsibility to either show us with power point, or write on the board the main issues, equations, figures, etc. covered in the chapter. Your goal is to make sure that we fully cover the material and that collectively we figure it out.

 

   Class Discussion

You are required to read two chapters every week. In class participation is expected, with everyone providing comments during every class.

 

   Class Links

·       Downloadable R code from the book

·       R studio

·       Cory’s notes on MCMC using a Mac with JAGS

o   BertTwoBugs

o   BernBetaBugsFull

o   PlotChains

o   FiconBrugs

o   ToyModelComp

o   HotHand

o   SystemsBrugs

o   SystemsJAGS

o   YmetricXsingleBrugs

o   SimpleRobustRegressionBrugs

o   SimpleLinearRegressionRepeatedBrugs

o   MultipleLinearRegressionBrugs

o   MultiLinRegressHyperBrugs

o   MultiLinRegressInterBrugs

o   ANOVAonewayRJAGS

o   LogisticOnewayAnovaHeteroVarBrugs

o   LogisticOnewayAnovaBrugs

o   MultipleLogisitcRegressionBrugs

o   PoissonExponentialBrugs

·       Rather than modifying your code to make it compatible with JAGS, you could try FakeBRugs

·       A new article investigating the many ways to cheat with NHST


Schedule of Class Meetings

 

 

Date

Chapters

Discussion Led By

Sep. 28

Introduction

 

Oct. 5

Chapters 1+2 (no one leads)

Chapter 3: Probability

Chapter 4: Bayes’ Rule

 

Carson

Mallorie

Oct. 12

Chapters 5+6: Binomial Distribution

Chapter 7: Metropolis Algorithm

Esther

Jordan

Oct. 19

Chapter 8: Gibbs Sampling

Chapter 9: Hierarchical Prior

Rachel

Jon

Oct. 26

Chapters 10: Model Comparison

Chapter 11: NHST

Liz

Nicole

Nov. 2

Chapter 12: Bayesian Null Hypothesis

Chapter 13: Bayesian Power Analysis

Evan

Andy

Nov. 9

Chapter 14: the GLM

Chapter 15: Bayesian One Sample t-test

Megan

Bernhard

Nov. 16

Chapter 16: Bayesian Regression

Chapter 17: Bayesian Multiple Regression

Cara

Dave

Monday

Nov. 21 (5-7pm)

Chapter 18: Bayesian Oneway ANOVA

Chapter 19: Bayesian Multifactor ANOVA

Erik

Sirawaj (Sean)

Nov. 30

Chapter 20: Bayesian Logistic Regression

Chapter 22: Bayesian Chi-Square

Kevin

Cory