Desingning social surveys effective for multilevel analysis

2022 Fiscal Year Final Research Report

• Project/Area Number: 17K18599
• Research Category: Grant-in-Aid for Challenging Research (Exploratory)
• Allocation Type: Multi-year Fund
• Research Field: Sociology and related fields
• Research Institution: The Institute of Statistical Mathematics
• Principal Investigator: Maeda Tadahiko 統計数理研究所, データ科学研究系, 准教授 (10247257)
• Project Period (FY): 2017-06-30 – 2023-03-31
• Keywords: 二段抽出 / 地点間異質性 / 級内相関 / 層化 / 日本人の国民性調査 / ウェブ調査

Outline of Final Research Achievements

The purpose of this study was to develop a survey design method, mainly from the perspective of sample design, that would be advantageous in some sense when employing multilevel analysis in the analysis, assuming a stratified two-stage sampling setting, which is a typical sample design in social surveys. For a multilevel analysis in this setting, the existence of variables with high intraclass correlation (heterogeneity among survey sites) is the key to a successful analysis. Although we have conducted our study based on the above-mentioned perspective, we could not achieve our initial goal because it is difficult to actively control for the variable (the content of survey items) from the viewpoint of sample design, mainly because inter-sites heterogeneity is a property of the variable (the content of survey items). On the other hand, empirical rules concerning the content of variables with high interaclass heterogeneity have been accumulated.

Free Research Field

統計科学,社会調査法

Academic Significance and Societal Importance of the Research Achievements

地点数がある水準以上に確保される必要がある全国レベルでの社会調査の設計においては，標本設計を変えることによってマルチレベル分析に対して有利（したがって別の設計を採った時の相対的な不利）が生じるという形での地点間異質性の制御が行えるわけではなく，基本的には項目に備わった性質として異質性が高いものについては，マルチレベル分析が行いやすくなる，といった結論を示したことが，消極的な結論ながら本研究の意義である。全国規模調査での代表性を担保するための設計を放棄して，極端に異質性が高くなる（ことが事前に分かっている）地域のセットを事前に指定した有意抽出以外の設定は困難であろうとの見通しが立った。