The Official Statistics Section of the SSA and the Australian Bureau of Statistics are pleased to present the following webinar
Advances in Data Linking Methodology at the ABS
with Daniel Elazar, Director of the Data Linking Methodology Section, Australian Bureau of Statistics
on Monday, 21 September 2020, at 1pm AEST (11am NZ time).
About this webinar
Data Linking methodologies provide a very useful way of combining data from two or more sources. Following the recent completion of the three year government funded Data Integration Partnership Australia initiative, ABS now has a highly developed suite of data linking methodological capabilities, infrastructure and has produced a wide range of linked data assets. This is greatly helping policy makers and researchers gain a much better understanding of Australian families, communities, industry, and the economy. This understanding is helping to improve the efficiency and delivery of government services in areas such as health, education, infrastructure, and other community services. At the same time, linked data assets are becoming more embedded in the statistical operations of many national statistical agencies, as a way to enhance or repair official statistics. In addition, there is a rapid expansion in the suite of statistical products offered to analysts by statistical agencies.
The ABS is one of several Accredited Integrating Authorities (AIA) in Australia which are approved under a Commonwealth initiative to undertake linkage of sensitive data for other organisations. Like the other AIAs, the ABS takesdata security and confidentiality very seriously and the protection of privacy and confidentiality is a top priority that extends to data integration activities: ABS has strong legislative framework, policies and systems safeguards and processes to protect all the data it holds.
This talk is aimed at research analysts using linked data and will cover some of the methods used at the ABS to link data, how we measure and assure the quality of the linked data products and the research work we are engaged in to improve these methods. In particular, I will cover the main differences between serial deterministic linking (the main production method we use) and Fellegi-Sunter probabilistic linking but argue why they are conceptually equivalent. There will also be a brief coverage of the person linkage spine we developed and it’s benefits. From an analytical perspective, I will cover what researchers need to be aware of when analysing linked data and some of the methods currently available for adjusting their analyses to account for the statistical uncertainty induced by the linkage mechanism.
About the presenter
Daniel Elazar has worked in Methodology Division at the Australian Bureau of Statistics for slightly over thirty years. He is currently the Director of the Data Linking Methodology Section whose major role is to develop, adapt and apply existing or new statistical methodologies related to record linkage. Previously, Daniel has worked extensively in the areas of statistical confidentiality, statistical applications for novel data sources, small area estimation and household and business survey methodology. Recently, Daniel delivered the 2019 SSA Ken Foreman Lecture on “Methodological Challenges in Balancing Data Confidentiality and Utility in an Increasingly Contested Data Environment”.
In his spare time, Daniel enjoys catching up with family and friends, gardening, going for long walks, crowd dis-pleasing movies and endures never ending home repairs.
This is a free event for members of SSA, employees of the ABS and NZSA members, but you will need to register.
To register, please click here.