Bayesian Data Analysis in SDT
SDTにおけるベイズ的データ分析
Yasuharu Okamoto
岡本安晴
Japan Women’s University
Abstract
Bayesian data analyses in equal- or unequal-variance Guassian signal detection theory (SDT) are proposed. Lee (2008) proposed a Bayesian analysis in SDT, but his method is for a simple equal-variance SDT with two response categories, yes and no, and uses a Markov chain Monte Carlo (MCMC) method, which requires some experience by a user. Before Lee’s proposal, Okamoto (2005) had proposed a simple grid method of Bayesian analysis in the simple SDT, and Okamoto (2006) had proposed an MCMC method for an unequal-variance SDT model with more than two response categories. The grid method by Okamoto (2005) does not require a special experience and a user can get an appropriate result of Bayesian analysis by adjusting values displayed on the form window. The MCMC method by Okamoto (2006) can be applied to data on SDT, which may have unequal-variances. Evidences of unequal-variance are presented in Green and Swets (1966). Both of Okamoto’s methods, which were presented in Japanese, are explained here in English.
Full text (pdf file): Click here
Full text (pdf file at CiNii): Click here
Programs
For equal-variance SDT model: Source code files (SDTSimple.zip) Release version (SDTSimple.exe)/.NET Framework 3.5 How to use manual (mansdtsimple.pdf)
For unequal-variance SDT model: Source code files (SDTRating.zip) Release version (SDTRating.exe)/.NET Framework 3.5 How to use manual (mansdtrating.pdf)