Statistical and Computational Models in Psychological Science - A/Prof Dan Navarro

What is this research about?

How should psychologists think about our data? In any real research situation, our primary goal as scientists is to use the data from our experiments to arrive at conclusions about the world. These conclusions necessarily require the researcher to make an inductive leap, generalising from the observations they have made. This presents a philosophical puzzle, because induction is fundamentally an ill-posed problem, and it is never possible to draw “safe” conclusions without relying on some untested (or untestable) assumptions. However, it also poses practical problems: what kind of statistical assumptions are sensible in a psychological context? Relatedly, are our data analysis tools fit for purpose, or should we as a discipline be developing new ones? Should researchers rely on “generic” statistical tests, or should we seek to build “specific” models for a task? These and other questions are investigated in this project.

Publications relating to this project:

S Tauber, DJ Navarro, A Perfors and M Steyvers (in press). Bayesian models of cognition revisited: Setting optimality aside and letting data drive psychological theory. Psychological Review

L Kennedy, DJ Navarro, A Perfors and N Briggs (in press). Not every credible interval is credible: On the importance of robust methods in Bayesian data analysis. Behavior Research Methods

MA Pitt, W Kim, DJ Navarro and JI Myung (2006). Global model analysis by parameter space partitioningPsychological Review, 113, 57-83


UNSW Computational Cognitive Science Lab