|Page tools: Print Page Print All|
Update from the Data Linking Team
The data linking team in the Analytical Services Unit (ASU) provide methodological support to a number of internal clients conducting data linking projects. These projects include the linking of a 5% sample of 2006 Census records to the 2011 Census to form the first wave of the Statistical Longitudinal Census Dataset (SLCD), the linking of deaths registrations to the 2011 Census to investigate Indigenous mortality, the linking of school enrolments data to the 2011 Census and the linking of the Migrants Settlements Database to the 2011 Census.
Research performed by data linking team members has resulted in two significant methodological improvements that have been implemented for the 2011 linking projects. The first of these is the use of the Expectation-Maximisation (EM) Algorithm to estimate the parameters of the models used for each linking run. The second is the use of sampling methods to determine the weight cut-offs for acceptance and rejections of record pairs as links. The data linking team believes that these methodological improvements will result in increased accuracy of the linked data compared to similar linking projects which were conducted with the 2006 Census data.
Linking projects which use name and address information as linking variables are referred to as Gold Standard projects. All name and address information from the Census is destroyed at the end of the Census processing period. For this reason, these projects need to be completed before the Data Processing Centre (DPC) shuts down in early December. The data linking team and our client areas are all currently very busy working to this deadline. There are also linking projects which do not use name or address information. These are known as Bronze Standard, and will continue after the DPC shuts down. Once all the linking has been performed, the data linking team will move on to the analysis of the linked data sets.
These documents will be presented in a new window.