A seminar by Dr Daniel Elazar - Australian Bureau of Statistics
Title: Practical approaches to accounting for statistical uncertainty when analysing linked data
Abstract:
The number of linked data products available to researchers and analysts has been growing rapidly over the last decade and will continue to grow with government funded initiatives such as the National Disability Data Asset (NDDA), a project led by the Australian Government Department of Social Services. NDDA will help bring together de-identified data from Australian, state and territory government agencies for approved research and analysis purposes. Linked survey and administrative data is invaluable for researchers and policy analysts to derive more accurate and detailed inferences about Australia’s social, health, welfare, and economic landscape. While linked data is being used more and more for a wider range of analyses, the application of practical and reliable approaches to account for the impact of linkage uncertainty upon inferences remains rare and elusive. Nevertheless, it is still fundamentally important to understand, measure and adjust for linkage uncertainty when we analyse linked data.
In this seminar, Daniel will give a brief introduction to data linking methods, outline some of the sources of linkage error that can arise, explore some new potential methods for quantifying linkage uncertainty and discuss some of the available methods for incorporating them in regression estimates. Although the approaches outlined here have not yet been fully tested empirically, I am hoping it will provide a stepping stone for others to research, evaluate, improve and extend these approaches into the future. The methods outlined in this talk can be viewed as providing the probabilistic linkage uncertainty matrices as input to methods covered in works such as Zhang and Chambers (Eds.) (2019) Analysis of Linked Data.
For further information, please contact RSFAS Seminars.
All information collected by the University is governed by the ANU Privacy Policy.