Informatics Methodology The Transition from GEIS to Banff
Informatics & Methodology The Transition from GEIS to Banff Chris Mohl, Yves Deguire, Rob Kozak, Chantal Marquis Statistics Canada Statistics Statistique Canada
Similarities of GEIS and Banff v Primarily for imputation of numeric data (nonnegative) v Modules for edit analysis, outlier detection, error localization and imputation v Fellegi-Holt rule of minimum change and Chernikova’s algorithm used to identify fields to be imputed v Edits must be expressed in linear form Informatics and Methodology 2 Generalized
Differences of GEIS and Banff GEIS Banff v Oracle/SQL v SAS v UNIX, mainframe v PC, UNIX, mainframe v Run modules in a specific order v Decoupled modules (run in any order) v User must write code v User can write SAS code, use SAS wizards or Banff processor (SAS code generator) Informatics and Methodology 3 Generalized
Methodological Enhancements in Banff v Estimation procedures can now handle negative data v Random error term can be added to imputation for all estimator functions v Additional options for outlier detection v More edit variables can be processed efficiently Informatics and Methodology 4 Generalized
Future Enhancements to Banff Version 2 (autumn 2005) v Negative data allowed in all modules v SAS wizards available for use Future versions? v Allow non-linear edits via Logiplus software v Functionality to allow use of qualitative data v Generalize the Banff processor Informatics and Methodology 5 Generalized
Demonstration of Banff Wednesday May 18 at 14: 00 11 th floor R. H. Coats Building Informatics and Methodology 6 Generalized
- Slides: 6