Using the Multi-level, Integrated Database Approach (MIDA) to Reduce Non-Response Error and Contextualize Survey Analysis

调查的回复率一直在下降，人们越来越担心无反应偏差正在增加，并破坏调查研究数据的质量。多层次的综合数据库方法（MIDA）寻求解决和纠正这个问题。MIDA通过以下方式极大地丰富了现有的样本框架：a)从样本框架中提取所有可能的信息用于数据收集和分析；b)通过从广泛的公共数据库中链接案例级和聚合级数据来增强样本框架；c)通过为受访者和非受访者记录数据收集过程中产生的大量数据来进一步加强数据收集。扩充样本框架中的数据可用于：a)通过向采访者提供更多关于目标样本的信息来协助数据收集；b)通过比较被调查者和非被调查者在扩充样本框架中已知的所有案例的广泛案例级和总体级信息来评估非反应性偏见；c)通过使用相同的情境数据来分析已完成的案例，从而进行情境化分析。MIDA的初步测试将从国家意见研究中心的国家样本框架中抽取400个地址样本，并用来自广泛公共数据源的案例级和汇总级数据来增加这些案例。然后，它将评估添加信息的完整性和准确性，例如通过跨信息源交叉检查数据。调查研究是研究人口最广泛使用的工具，但回复率的下降引起了人们的担忧，即由于非反应偏差的增加，调查数据的质量正在下降。MIDA具有以下潜力：a)通过在数据收集过程中为采访者提供更多有用的信息来提高回复率；b)更好地评估调查中存在的非反应偏差水平，然后制定适当的权重来调整检测到的任何偏差；c)通过将上下文、地理数据自动作为每个调查数据文件的一部分来改进数据分析。因此，MIDA应使科学研究从调查中得到普遍改进。

Response rates in surveys have been declining, and there is growing concern that non-response bias is increasing and undermining the quality of survey-research data. The multi-level, integrated database approach (MIDA) seeks to address and redress this problem. MIDA greatly enriches existing sample frames by a) extracting all possible information from the sample frame for use in data collection and analysis, b) augmenting the sample frame by linking in both case-level and aggregate-level data from a wide range of public databases, and c) further enhancing the data collection by recording for both respondents and non-respondents a large degree of paradata generated during the data-collection process itself. The data in the augmented sample frame can be used to a) assist in data collection by providing interviewers with more information about the target sample, b) assess non-response bias by comparing respondents and non-respondents on the extensive case-level and aggregate-level information known for all cases in the augmented sample frames, and c) contextualize analysis by using the same contextual data for analysis on the completed cases. This initial test of MIDA will draw a sample of 400 addresses from the National Opinion Research Center national sample frame and augment those cases with case-level and aggregate-level data from a wide range of public data sources. It will then assess the completeness and accuracy of the added information such as by crosschecking data across information sources.Survey research is the most widely used tool for studying human populations, but the decline in response rates has raised the concern that the quality of data from surveys has been falling due to increased non-response bias. MIDA has the potential to a) improve response rates by providing interviewers with more useful information during data collection, b) better asses the level of non-response bias that exists in surveys and then develop appropriate weights to adjust for whatever bias is detected, and c) improve data analysis by making contextual, geographical data automatically available as part of every survey data file. MIDA should thus lead to a general improvement in scientific research from surveys.