Parallelization of statistical estimations for psychometric models and its application to large scale data

2014 Fiscal Year Final Research Report

• Project/Area Number: 23530968
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Research Field: Experimental psychology
• Research Institution: The National Center for University Entrance Examinations
• Principal Investigator: OTSU Tatsuo 独立行政法人大学入試センター, 研究開発部, 教授 (10203829)
• Co-Investigator(Kenkyū-buntansha): MIYANO Hisao 大学入試センター, 研究開発部, 教授 (90200196)
• Project Period (FY): 2011-04-28 – 2015-03-31
• Keywords: 潜在変数 / 並列計算 / テスト理論 / 因子分析 / スプライン / テスト等化

Outline of Final Research Achievements

The role of multivariate statistical analyses has increased in psychology and related social and behavioral sciences. Extensions of their application area led to enlargement of target data and computational sizes. We developed parallelization for speeding up estimations in statistical methods for psychometric models, especially nonlinear latent variable models. We applied the developed method to analyses of latent structures in educational testing. The parallelization of non-linear factor analysis on six cores PC make the speed of estimation more than three times faster compared to serial computation.

Free Research Field

計量心理学