Statistical data editing near the source using cloud computing concepts George Pongas, Christine Wirtz -Eurostat
Editing near the source Accelerates speed of final delivery to users and institutions Checks and imputations are near the respondent Data knowledge is frequently more profound in the primary collector institutions Logical proximity is better than physical: Data and application sharing
Cloud and SOA in few Lines Separates ownership and usage of data storage computer power and application development and execution (cloud) Cloud variants are IaaS, PaaS, SaaS Cloud architectures are: Public Private Mixed Community Based on web technologies and independent software components to interlink on demand (SOA)
Data Editing in Eurostat High volume of arrivals (>60.000 per year) Format heterogeneity Data checking absorbs substantial volume of human resources Erroneous data imply communications with MS Eurostat as a rule does not Impute… Interest to have a Common distributed solutions
Eurostat’s web enabled system for editing (Editing building block (Ebb) Completely Metadata Driven Exists in 2 versions: PC version Web-based version Technologies used: ANTLR Java Tomcat or Weblogic Hibernate Postgres or Oracle
EBB Information Flow
Implementation Details EBB is written using a set of Web services of the following types: Administration Program Job
EBB functionalities Support of categorical, text and numeric variables Separation of programmer and user interfaces Conditional and unconditional rules Multi-record rules Deterministic imputation Use of auxiliary data File operations Special functions (unicity, duplication checks ...) Outliers (HB, Sigma Gap, Terror) Input/output of data/metadata Reporting
Usage until now Embedded in SAS (for microdata editing) To distribute to data providers as standalone version FDI (foreign direct investments) ITS (international trade in services) SBS (structural business statistics) CVTS (continuous vocational training survey), AES (adult education survey)