Analysis
Filter results by
Search HelpKeyword(s)
Author(s)
Survey or statistical program
Results
All (3)
All (3) ((3 results))
- Articles and reports: 11-522-X202500100019Description: Accurate and efficient record linkage is crucial for maintaining a comprehensive and current Statistical Business Register (SBR) at Statistics Canada. Linking external business lists to the SBR by name presents computational and methodological challenges, especially as data volumes grow. This paper describes a scalable methodology that employs blocking techniques to constrain the computational search space and integrates multiple similarity measures—from edit distances and n-gram overlaps to embedding-based methods using Sentence-BERT (SBERT)—to identify likely matches. By combining simple character-level comparisons with more advanced semantic embedding methods, the approach can adapt to various naming conventions and complexities. While it does not guarantee superior accuracy in all circumstances, it offers a pragmatic balance between computational feasibility and linkage quality.Release date: 2025-09-08
- Articles and reports: 12-001-X201800254957Description:
When a linear imputation method is used to correct non-response based on certain assumptions, total variance can be assigned to non-responding units. Linear imputation is not as limited as it seems, given that the most common methods – ratio, donor, mean and auxiliary value imputation – are all linear imputation methods. We will discuss the inference framework and the unit-level decomposition of variance due to non-response. Simulation results will also be presented. This decomposition can be used to prioritize non-response follow-up or manual corrections, or simply to guide data analysis.
Release date: 2018-12-20 - 3. Getting around a conceptual difference between two data sources when producing estimates ArchivedArticles and reports: 11-522-X20050019481Description:
The Survey on Employment, Payrolls and Hours is a monthly survey using two data sources: a census of administrative records and an establishment survey. The survey data is used to build models in order to mass impute several derived variables on the administrative source. The survey design relies on the fact that the concepts for number of employees and gross monthly payroll are the same on the two data sources. In this presentation, we will describe different solutions that were brought to the survey design and to the mass imputation model to allow us to get around this conceptual difference, hence producing estimates that are more stable in time. Results from different estimation scenarios for average weekly earnings will be given to conclude the presentation.
Release date: 2007-03-02
Articles and reports (3)
Articles and reports (3) ((3 results))
- Articles and reports: 11-522-X202500100019Description: Accurate and efficient record linkage is crucial for maintaining a comprehensive and current Statistical Business Register (SBR) at Statistics Canada. Linking external business lists to the SBR by name presents computational and methodological challenges, especially as data volumes grow. This paper describes a scalable methodology that employs blocking techniques to constrain the computational search space and integrates multiple similarity measures—from edit distances and n-gram overlaps to embedding-based methods using Sentence-BERT (SBERT)—to identify likely matches. By combining simple character-level comparisons with more advanced semantic embedding methods, the approach can adapt to various naming conventions and complexities. While it does not guarantee superior accuracy in all circumstances, it offers a pragmatic balance between computational feasibility and linkage quality.Release date: 2025-09-08
- Articles and reports: 12-001-X201800254957Description:
When a linear imputation method is used to correct non-response based on certain assumptions, total variance can be assigned to non-responding units. Linear imputation is not as limited as it seems, given that the most common methods – ratio, donor, mean and auxiliary value imputation – are all linear imputation methods. We will discuss the inference framework and the unit-level decomposition of variance due to non-response. Simulation results will also be presented. This decomposition can be used to prioritize non-response follow-up or manual corrections, or simply to guide data analysis.
Release date: 2018-12-20 - 3. Getting around a conceptual difference between two data sources when producing estimates ArchivedArticles and reports: 11-522-X20050019481Description:
The Survey on Employment, Payrolls and Hours is a monthly survey using two data sources: a census of administrative records and an establishment survey. The survey data is used to build models in order to mass impute several derived variables on the administrative source. The survey design relies on the fact that the concepts for number of employees and gross monthly payroll are the same on the two data sources. In this presentation, we will describe different solutions that were brought to the survey design and to the mass imputation model to allow us to get around this conceptual difference, hence producing estimates that are more stable in time. Results from different estimation scenarios for average weekly earnings will be given to conclude the presentation.
Release date: 2007-03-02