Stochastic Model approaches to the analysis of social survey data with missing values

2023 Fiscal Year Final Research Report

• Project/Area Number: 20K02171
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Review Section: Basic Section 08010:Sociology-related
• Research Institution: Ritsumeikan University
• Principal Investigator: Nakai Miki 立命館大学, 産業社会学部, 教授 (00241282)
• Project Period (FY): 2020-04-01 – 2024-03-31
• Keywords: 縦断データ / 欠損値 / 潜在クラス分析 / 主観的幸福感 / 隠れマルコフモデル

Outline of Final Research Achievements

During the past four years, the project focused on applying stochastic model to analyzing sociological data, especially longitudinal data, with missing values. We obtained new sociological findings by applying some stochastic models. The main results are:
1) First, we considered appropriate parameter estimation for models when dealing with longitudinal survey data with missing values. We propose a hidden Markov model to analyze longitudinal data. Parameters of the latent model are: Initial and transition probabilities are parameterized according to a multinomial logit model. Model inference is carried out through maximum likelihood with the Expectation-Maximization algorithm.
2) Second, in order to improve the method and verify its effectiveness, we applied the model to social survey data and analyzed the transition process of variables (latent variables) that are not directly observed.

Free Research Field

社会学

Academic Significance and Societal Importance of the Research Achievements

近年の調査データの統計解析技法の顕著な発展に比して、欠損・欠測を含む社会調査データへの対処手法はあまり注意が払われてこなかった。本研究の成果は推定バイアスを回避するための技法上の洗練という点で重要な学術的意義を持つ。また、学際的・国際的な共同研究を進め議論を深めることを通じて、日本の社会意識態度のパターン分類とその規定構造や主観的幸福感の経時的変化を明らかにしたことが社会的意義である。