Research Seminar
Bayesian methods in psychology: An introduction and some applicationsJeff Rouder & Eric-Jan WagenmakersUniversity of Missouri-Columbia and University of Amsterdam | |
| Abstract: | PART 1: Using Bayesian hierarchical models to search for structure in recognition memory (Jeff Rouder) Science has conventionally progressed by identifying regularities or invariances; for instance, planets orbit the sun as ellipses. These invariances then serve as the phenomenon to be explained by theory; for instance, ellipitical orbits may be explained by the Newtonian theory of mechanics. In this mode, the search for invariances is a necessary precursor to theoretical development. Though this approach has been fruitful, it has not been adopted by psychologists, who tend to focus on verbal explainations of differences and idiosyncracies across conditions and populations. The search for invariances is complicated in many domains by the presence of unintended or nuisance variability. For instance, in memory research, there is variability from the selection of participants and items in addition to that from the mnemonic processes at hand. The conventional approach is to aggregate across these nuisance sources, but this aggregation may distort the stucture and mask important invariances in the data. To address this problem, my collegeagues and I have developed a collection of Bayesian hierarchical models for modeling nuisance variability so that structure in psychological processes may be explored. I apply these models to recognition memory, where there is a vigorous debate whether memory is supported by a single mnemonic process or by multiple distinct processes. Model analysis reveals that ROC curves across people and conditions form an orderly field, much like the order in gravitational of magnetic fields. As a consequence, the structure of recognition memory seems to be one-dimensional and may be accounted for with a single parameter of mnemonic strength. PART 2: Bayesian Hypothesis Testing For Dummies (Eric-Jan Wagenmakers) In the field of cognitive psychology, the p-value hypothesis test has established a stranglehold on statistical reporting. This is unfortunate, as the p-value provides at best a rough estimate of the evidence that the data provide for the presence of an experimental effect. An alternative and arguably more appropriate measure of evidence is conveyed by the Bayesian hypothesis test, which prefers the model with the highest average likelihood. One of the main problems with the Bayesian hypothesis test, however, is that it often requires relatively sophisticated numerical methods for its computation. Here we draw attention to what is probably the simplest exact solution to the computation of the Bayesian hypothesis test. The solution, known as the Savage-Dickey density ratio is valid for nested models and under certain plausible restrictions on the parameter priors. Practical examples demonstrate the method's validity, generality, and flexibility. PLEASE NOTE THE UNUSUAL DAY, TIME AND LENGTH OF THE SEMINAR. |
| Date: | Thu May 14, 1:00 pm - 3:00 pm |
| Place: | room 02.51 (Department of Psychology, Tiensestraat 102, 3000 Leuven) |