講演者のシアトルのワシントン大学John Miyamoto博士は、長年数理心理学に貢献され、順位依存期待効用モデルの功績で、Society for Mathematical PsychologyのBest Paper Awardを受賞しています。このたび，東工大の山岸侯彦先生の招聘により来日し，京大にお招きしました。
John Miyamoto博士は，現在，因果関係のメンタルモデルに関するCausal Bayes Netsの研究を進めています。これは以前講演会をおこなったSloman博士の研究とも関わります。今回の講演は，ベイス統計学的手法の最近の展開と統計教育や心理学教育にどのように活かすかという内容で した．
TITLE: Is Bayesian Statistics About to Become the Standard Methodological Framework of Research Psychologists?
講演者: John M. Miyamoto ワシントン大学心理学部
Recent developments in Bayesian statistical methodology make it possible to change the statistical methodology of a research-oriented psychology department from one based on classical statistics to one based on Bayesian statistical methods. In my talk, I will review the theoretical and computational developments that now make it possible for non-mathematical research psychologists to use Bayesian statistics in teaching and research, and I will argue that a shift in paradigm from classical statistics to Bayesian statistics will greatly benefit the research activities of psychologists.
The standard statistics curriculum as taught in an American graduate psychology program is based on a frequentist interpretation of probability, i.e., probabilities are the long-run relative frequencies of independently repeatable events. This approach places central importance on concepts like tests of statistical significance, the p-value of a test, confidence intervals, and measures of goodness of fit. Until recently, an alternative approach called Bayesian statistics was generally ignored within psychology programs. Of course, Bayesian statistics has received considerable attention from professional statisticians, and Bayesian ideas have exerted enormous influence in cognitive psychology and neuroscience, but the statistical training of the typical psychology program has remained firmly tied to the classical frequentist framework. I believe that we are now on the verge of a paradigm shift in the statistical methodology that is taught in a graduate psychology program.
Why will a paradigm shift towards Bayesian methods be good for psychology:
The theory and computational methods of Bayesian statistics have matured to the point where it is now possible for the non-mathematical psychologist to use Bayesian ideas in their thinking, and Bayesian statistics in their practical data analysis. I have three general arguments for why this shift will benefit psychology:
(i) strictly on theoretical grounds, Bayesian statistical methods are better than classical statistical methods;
(ii) Bayesian ideas will help psychologists understand the behavior of organisms in uncertain environments (this is not a new idea to students of judgment and decision making or, for that matter, of animal cognition);
(iii) Bayesian models play a major role in neuroscience and a Bayesian emphasis in the standard statistics curriculum will help students understand these models. These advantages should accrue to psychologists in every specialty.