Wednesday 23rd November 2022, 15:00–16:30 (Australia/Melbourne), Zoom Breakout Room 3
Sample surveys are a common way to collect data and make inferences about a population of interest. Stakeholders and users of the data often need to evaluate or inform policies at the local area level, and additional requirements may later arise that were unknown at the survey design stage. Cost considerations necessarily limit the number of surveys that can be collected at the local level, so that standard survey estimation techniques often yield unreliable results when applied to local areas (such as Local Government Areas, LGAs).
“Small area” statistical methods have been developed for these situations, where the survey data alone is too scarce to yield reliable estimates. The methods work by combining person-level data (from the survey) with area-level auxiliary data (from the Australian Census of Population and Housing, for instance) to improve the reliability of estimates. The survey data is used to calculate the contemporary, quantitative relationships between respondent characteristics and the survey’s key outcome variables. These relationships are then “projected” onto each individual area, using the area-level auxiliary data, to generate reliable estimates that reflect the population characteristics of each area. This approach makes maximum use of the available survey and auxiliary data and ensures that results for each area are as reliable and representative as possible.
This workshop will give an overview of considerations, methods, tools and outputs for small area estimation. Topics covered will include the following:
* Design considerations, including questionnaire items and sources of auxiliary data;
* Preparation of survey and auxiliary data for modelling;
* Model building and production of estimates; and
* Options for presentation and visualisation of results.
Material and examples will focus on models for small areas, but mention will also be made of small groups in the population to which the same techniques can be applied. Comments will be made about implementation of the methods in R, including package recommendations, but references to the applicable literature will be provided to enable working with other software.
The target audiences for the workshop are policy makers and quantitative researchers who have an interest in getting more detailed results from their surveys, along with survey analysts wishing to expand their statistical capabilities and tools.
Andrew has almost 20 years of experience as a statistician and quantitative researcher, with particular expertise in survey research, analysis and reporting. His interests and capabilities include cross-sectional and longitudinal surveys, survey weighting and estimation, psychometrics and small area methods.
He has completed a Bachelor of Applied Science in mathematics and statistics at the Queensland Institute of Technology, a Diploma in Education at The University of Queensland, and a Master of Applied Science by research and thesis at the Queensland University of Technology. He is an Accredited Statistician and full member of the Statistical Society of Australia.