Congratulations to the winners of the 2019 IAOS Young Statisticians Award!
We thought you might want to learn know more about each of the winners. You will find their profiles, a summary of each paper, and each winner’s response to a question relating to their work or career here.
Nancy Torrieri - Interview Editor
James Farnell is a data and research scientist with experience in statistical methodology and data visualization. He is currently employed by the Australian Bureau of Statistics where he is focused on contingency planning, content testing, and developing administrative data-based methodologies for the 2021 Census. Earlier in his career he worked as a commercial research scientist on solar photovoltaics. James has a Bachelor of Science degree from Australian National University and has completed a variety of methodological courses with the Australian Bureau of Statistics.
Peta Darby is currently an Assistant Director at the Australian Bureau of Statistics. Peta has worked in a variety of statistical and strategic roles at the Australian Bureau of Statistics. Her recent focus has been on understanding the population coverage and potential statistical applications of a range of administrative data sources. Peta has a Bachelor of Statistics from the Australian National University.
The prizewinning paper written by James and Peta, Administrative data-informed donor imputation in the Australian Census of Population and Housing, focuses on donor imputation and its application within the Australian Bureau of Statistics to address unit non-response in the Australian Census of Population and Housing. Census responses for private dwellings are typically submitted at a dwelling level containing all people present on Census night. In the 2016 Census, non-responding occupied units were treated by donor imputation. The 2016 Post Enumeration Survey (PES) indicated that the age distribution of the imputed population was skewed compared to that of the non-responding population. This indicated that the existing method of donor selection could be improved with enhanced selection information.
The paper describes the status of administrative data in Australia and the authors’ proposal to improve the 2016 Census imputation approach. The Administrative Data Informed Donor Imputation (ADDI) method incorporates administrative data to inform the choice of donors in donor imputation. Similar to the 2016 Census imputation methods, the ADDI follows a nearest neighbor approach in combination with random hot deck imputation. The authors’ paper briefly describes the five main steps to implement the ADDI method, then explores further the implementation of this method in subsequent sections of the paper.
The ADDI method was implemented to re-impute non-responding units in the 2016 Census. An analysis of the results show that the imputation results based on the ADDI method match the PES estimates more closely than the 2016 Census imputation method.
In the paper, the authors propose further research to refine the methods they used. They also expressed the view that expansion of the range and coverage of administrative data sources may improve results. Implementation of the ADDI method in combination with more direct methods such as administrative data substitution also may be useful.
We asked James to describe the challenges he foresees in the implementation of the ADDI method in the broad context of the future development of methods to improve the use of administrative data. James stated that the ADDI method has already been developed for use in the 2021 Census. It forms one part of a broader deployment of administrative data to improve Census enumeration, processing and contingency planning. The successful demonstration of these benefits in the Census main event may spur further investment in innovative uses of existing data sets. A major consideration is ensuring the public is brought along with developments on how their data are used, and how these uses benefit Australia.
To read the entire paper, please visit: http://bit.ly/Farnell-Darby.