Surveys and Data Science at Statistics Canada

• Activity Number: 507 - How Can Data Science Improve Surveys?
• Type: Topic Contributed
• Date/Time: Thursday, August 11, 2022 : 8:30 AM to 10:20 AM
• Sponsor: Survey Research Methods Section
• Author(s): Eric Rancourt*
• Companies: Statistics Canada
• Keywords: Administrative Data; Big Data; Inference; Machine Learning; Non-Probabilistic

Abstract

Statistics Canada is mandated to produce information on a variety of topics about social, economic and environmental aspects of the Canadian society. This task amounts to a large number of statistical programs that are almost all based on surveys. These programs have evolved over the years, making use of a wide array of methods from the survey sampling theory. With the advent of data science, there is now a large new set of techniques that have emerged in recent years and that have started to be used in statistical programs. This presentation highlights some of the recent tests and implementations of data science techniques in statistical programs at Statistics Canada and points out some of the main successes and challenges.