The Centre for Sample Survey Methodology (CSSM) undertakes fundamental, grant supported and contract research, major consulting projects and professional training in statistical methodology for the design and analysis of sample surveys. Its mission is to provide international leadership in statistical theory and practice for sample surveys and censuses, through research, teaching, postgraduate supervision and partnerships with governments and industry.
Centre for Sample Survey Methodology
Director: Senior Professor David Steel
Phone: +61 2 4221 5976
Email: dsteel@uow.edu.au
Location: Building 39C Room 263
The CSSM is a centre of excellence in the design of data collection strategies for complex populations and their subsequent analysis. The Centre has international expertise in survey design and analysis; complex data analysis and estimation methods; small area statistics; privacy and confidentiality analysis; methodology for combining and analysing data from different sources, including administrative, linked and big data.
A distinctive feature of CSSM’s research is the focus on statistical design and analysis, which involves developing cost-effective methods for generating relevant and reliable data and accounting for the data generating processes in the analysis. Its research program considers the role of sample surveys and other data, including big data sources, and the application of sampling methods in the development of data science.
The research focus of CSSM builds on the fact that modern data requirements require moving away from the traditional concept of a sample survey as a free-standing information collection and analysis entity. Target populations are more dynamic, much less clearly defined (e.g. networks) and much harder to measure. Furthermore, conventional methods of sampling and data collection are rapidly becoming more expensive and technology is providing new ways of obtaining data. As a consequence, modern sampling design and analysis is evolving from its traditional emphasis on how selection is implemented and accounted for to how samples from quite different sources, and of varying methodologies, coverage and selection processes, can be readily integrated and analysed. Sampling inference is adapting to this new data collection paradigm, with the traditional focus on sampling error accompanied by increasing emphasis on how basic ideas of uncertainty and selectivity should be characterised in the resulting mix of factors that affect data quality and usefulness, including informative response and participation processes, non-response errors, linking errors, measurement errors and model specification.
Aim & research projects
The aim of CSSM is to ensure that sampling statisticians continue to make substantial contributions to how data are collected, analysed and interpreted in the modern environment of big data and data science, and ensuring that the value from the information in these data is maximised. In this context, CSSM currently has specific research projects in:
- Sample design for analysis
- Network autoregressive multilevel models
- Split questionnaire designs for data science.
- Survey design to estimate and adjust for the selection mechanism in non-probability-based data sources.