IASS Webinar 38: Data Integration, Data Linkage and Linked Data Analysis

Watch the webinar recording

Slides

Webinar Abstract

The increasing availability of multiple data sources to investigate complex socio-economic, agricultural, health and environmental phenomena represents a great opportunity for survey statisticians, given the need to reduce the cost of data collection and the increasing demand for detailed information. Combining already available data sources is a key element in complementing and increasing the relevant information available in a single source. However, sometimes lack of standardisation, different forms and structures (if any) of data pose challenges to data linkage procedures and may result in an error-prone linkage. Probabilistic data linking comprises a set of statistical methods for combining multiple sources, recognising the same units even in the absence of common identifiers. In addition, probabilistic linkage produces some measure of the potential errors generated by the statistical integration procedure itself. The propagation of uncertainty from data linkage to downstream analyses based on linked data is a topic of growing interest to survey statisticians, with some interesting proposals for adjusting standard methodology (and more).