The role of Statistical Computing in delivering quality

The role of Statistical Computing in delivering quality Amy Large Statistical Computing Branch Survey Methodology & Statistical Computing Division, Research, Development and Infrastructure Directorate

Presentation Outline • Background and context • Statistical Computing Projects • ONS Strategic Aims • Conclusions

Context (general) • Do we have common understanding? The error seems to have happened because the national rail operator gave the wrong dimensions to train company. "When you separate the rail operator from the train company, this is what happens. " Transport Minister, Frederic Cuvillier

Some costly software errors

Context (ONS) • Restructure April 2012 Methodology IT Research, Development and Infrastructure Directorate Statistical Computing Branch

Statistical Computing Branch Structure • Small team • Not linked to a specific business process • Centrally funded, with some funding for strategic projects • Multiple routes for engaging with project work • Hub and node working approach

Projects • Type A: involvement in large projects, act to help interpret requirements as a bridge between business areas and developers. • Type B: carry out small-scale development work, e. g. replacing spreadsheet processes. • How we work with each type of project will vary depending on requirements

ONS Strategic Aims • Part of the ONS Strategy (published March 2013) • Nine Aims: 1. Improve 2. 3. 4. 6. 7. 8. 9. 5. Have Inform Dramatically Be Keep Be Have at highly a the statistical skilled the flexible debate quality forefront data regarded and improve and we powerhouse and motivated efficient of hold have by minimise integrating the our secure acommunication greater processes customers people atthe and risk impact heart who exploiting of and forare errors of of onthe our decision statistics producing data Government enthusiastic systems fromfor making and multiple trustworthy statistical for Statistical analyses change sources production, statistics Service and underpinned the analyses thatsound European by anticipate methodology Statistical their needs System 6. Improve quality and minimise the risk of errors

Strategic Aim – Flexible and efficient processes • Understanding how to build a flexible system • • Hard coding / parameterisation Modular code Shared code Documentation and on-going support • Understanding how to build an efficient system: • Right software / platform / method • Knowing your data • Programming good practice

Strategic Aim – Sound Methodology • System redevelopment – quality assure and test against current processes • Modular code – ‘Plug and play’ data DS 1; set DS 1; %ratio_imputation; %near_neighbour_imputation; run; • Share common functionality

Strategic Aim – Improving quality • Statistical Quality comes from (ESS): • Output Quality: – Relevance, Accuracy, Timeliness & Punctuality, Accessibility & Clarity, Comparability, Coherence • Process Quality: – Efficiency, Flexibility, Transparency, Robustness, Effectiveness, Integration

Strategic Aim – Minimising the risk of errors • Common understanding • How do errors occur? • • Human Data Process Lack of QA • Project work often dictated by strategic review

Branch Objectives • Standard setting • Reduction of risk • Solution re-use • Appropriate solution selection

Conclusions • New way of working • More demand with increasing visibility and reputation • Successes: • Training – responsive, relevant, looking wider than the branch • Iterative, interactive, responsive – Zero Hours Contracts, Human Capital. . . • Reducing risk across the Office

John Pullinger – National Statistician “ mobilising the power of data to help Britain make better decisions about our future”

Contact Amy Large amy. large@ons. gsi. gov. uk Daniel Lewis (Branch Head) daniel. lewis@ons. gsi. gov. uk