Bayesian Data Analysis Colloquium Series
Dr. Kruschke is a Professor of Psychological & Brain Sciences and an Adjunct Professor of Statistics. Abstract: Suppose a researcher collects data from subjects in two groups. Question: Should the interpretation of the difference between groups depend on whether the researcher planned a fixed sample size in advance, or instead planned to sample until the end of the week, or instead planned to collect a larger sample but was unexpectedly interrupted? Should the interpretation of the difference between groups depend on whether the researcher intends to compare with other groups when their data become available next week? In classical statistics, the answer is yes: p values and confidence intervals depend on why data collection stopped and on what other comparisons might be made. Moreover, confidence intervals carry no distributional information and are only minimally useful for predicting new data. On the other hand, Bayesian data analysis does not suffer these problems. Bayesian analysis provides complete information about the relative credibilities of all candidate parameter values. Bayesian analysis applies seamlessly to small samples, large samples, unbalanced designs, missing data, outliers, etc. Bayesian analysis software is flexible and can be used for virtually any data-analytic model. And it's free. Bayesian Data Workshop Date: Saturday, October 22, 2011 Time: 9:00 a.m. - 12:00 p.m. Location: Stewart Center, Room 214 For details on the workshop (including software) and how to prepare (you may want to bring your own laptop), go to http://www.indiana.edu/~jkkteach/WorkshopPurdue2011.html ) |
