Likelihood-free Bayesian analysis of memory models.

• DOI: 10.1037/a0032458
• Publication Date: 2013-04-15T16:05:44Z
• Authors: Brandon M. Turner; Simon Dennis; Trisha Van Zandt
• Keywords: Approximate Bayesian Computation; Marginal likelihood

Many influential memory models are computational in the sense that their predictions derived through simulation. This means it is difficult or impossible to write down a probability distribution likelihood characterizes random behavior of data as function model's parameters. In turn, lack these cannot be directly fitted using traditional techniques. particular, standard Bayesian analyses such impossible. this article, we examine how new procedure called approximate computation (ABC), method for analysis circumvents evaluation likelihood, can used fit data. investigate bind cue decide model episodic (Dennis & Humphreys, 2001) and retrieving effectively from (Shiffrin Steyvers, 1997). We hierarchical versions each Dennis, Lee, Kinnell (2008) Dennis (2012). The ABC permits us explore relationships between parameters well evaluate relative fits data-analyses were not previously possible.