Traditionally, much of the data within the social sciences and market research were purposely collected. In both qualitative research, surveys or experiments, the researcher is in control of the design of data collection. Research designs align with the goals of the study, and are targeted towards answering the most important research question(s) the researcher has.

Nowadays, more and more data are not designed, but rather ‘found’. There is large variety in different kinds of found data: social media, online communication and (web) browsing data are often used in a variety of applications, but so are administrative data from governments, citizen science data, or publicly available (text) data from businesses for example. A key characteristic of found data is that the data were not collected with doing research in mind. Often, raw found data are not fit to answer a particular research question a researcher has. There is a need to process found data, (dis)aggregate them, and merge them with other datasets to be usable for research. Data integration is the process of merging or combining data from multiple sources to produce statistics.

One problem is that it is often unclear how to integrate data exactly. Integration methods are both research-question and data-source specific, making it difficult to establish a general methodology for how to integrate data. It also means that each time a statistic has to be computed, methods have to be (re)-evaluated making data integration a resource-intensive technique.

In this talk, Peter Lugtig will discuss how data-quality frameworks can be used to understand when and how to integrate data effectively. He will discuss several use cases, and show how already exisiting data quality frameworks can be used to understand how to integrate data for these specific use cases. Data quality frameworks can also be used in evaluating what datasources potentially would be useful when data integration is considered as a technique. He will also argue for the more frequent use of both designed data and found data within one study as a way to increase the quality of statistics produced by data integration.