Topic (i): Selective editing / macro editing

1
Topic (i) Selective editing / macro editing

• Discussants
• Orietta Luzi - Italian National Statistical
  Institute
• Rudi Seljak - Statistical Office of Slovenia
• Pedro Revilla - National Statistical Institute of
  Spain

2
Background

• The adoption of selective editing / macro editing
  approaches can significantly improve the
  efficiency of the data editing process
• This is especially true in the case of business
  surveys where usually a small amount of
  measurement errors have a decisive influence on
  the final results and therefore the appropriate
  usage of selective editing and macro editing
  techniques can significantly reduce the cost of
  the data editing process with no major impact on
  the quality of the final results
• Developing tools implementing selective editing
  models/functions may validly support further
  harmonization and widening use of efficient
  editing strategies in NSIs

3
Topic contents

• Practical implementation of local and global
  score functions
• Integration of selective editing / macro editing
  techniques within the statistical process
• Tools to support selective editing / macro
  editing, including visualization approaches
• Validation of selective editing strategies,
  inference issues
• Editing of unstructured data

4
Long Presentations

• Italy - Use of administrative data for selective
  editing the case of business investments
• Spain - Score Functions under the Optimization
  Approach
• United Kingdom Maintenance of Selective
  Editing in ONS Business Surveys

5
Short Presentations

• Sweden Experiences from Selective Data Editing
  at Statistics Sweden
• France Selective Editing Techniques and
  Seasonal Adjustment of Short-Term Economic
  Indicators
• France Using R-indicators to monitor household
  surveys and prioritize data collection an
  application to the 2010 Household Wealth survey
  in France
• Japan An Assessment of Automatic Editing via
  the Contamination Model and Multiple Imputation
• United States Text Analysis Tools for Editing
  and Verification
• Sweden Adjusting for Remaining Measurement
  Errors after Selective Editing

6
Short Presentations

• Sweden Adjusting for Remaining Measurement
  Errors after Selective Editing (T. Laitila, A.
  Norberg)
• Model-based approach for bias correction of
  estimators due to the remaining errors in
  selective edited data sets, where the observed
  measurement errors in the edited units are
  related to their scores.
• The estimated models are used to predict
  measurement errors in unedited units and their
  associated bias.
• Experimental application to data of the sample
  survey on establishments and their salary
  payments to employees carried on at Statistics
  Sweden.

7

• Topic (i) Selective editing / macro editing
• Discussion

8
Overall summary

• Session papers provide a wide-ranging overview of
  current research and development activities in
  the area of selective / macro editing carried out
  at various NSIs with the aim of enhancing the
  efficiency of survey editing strategies
• Theoretical formalizations of the selective
  editing paradigm are proposed, in a situation
  where no accepted theory has been developed in
  this area yet demanding modelling problems come
  out from presentations
• The need of a common theoretical evaluation
  framework for quality assessment of selectively
  edited data is an area requiring further
  research at also an international perspective
• Issues related to the development and
  dissemination inside NSIs of generalized
  IT-tools for selective editing are discussed,
  with specific attention to those relating to
  applicability, maintenance and optimization of
  the implemented methods/models
• New application contexts where selective / macro
  editing principles can contribute to most
  efficient statistical production processes are
  explored

9
Points for discussion

• Integrating selective / macro editing in existing
  EI process
• Advantages (increased efficiency, burden
  reduction,)
• Costs (re-design of processes, models/parameters
  maintenance/update due to changes of
  populations/information contexts, )
• Cultural issues (changes in attitude needed,
  importance of support by the survey staff,
  importance of constant training and education,)
• Countries experiences
• Papers from Italy, Japan, UK, Sweden (A. Norberg)

10
Points for discussion

• Formalization of selective editing
• Ad hoc methods VS theoretical frameworks
• How to estimate the model parameters?
• How do we apply the developed frameworks in
  practice?
• Papers from Italy and Spain

UNECE Work session on SDE, Paris, 28-30 April 2014
Topic (i) Selective editing / macro editing
10
11
Points for discussion

• Standard tools for selective editing
• Advantages (costs savings, harmonization, easy to
  document and to update, )
• Drawbacks (specificity against generalization,
  validity of models assumptions, black-box
  effect, .)
• International co-operation, sharing solutions
• SeleMix (papers from Italy, Japan), Selekt
  (papers from Sweden, UK)

12
Points for discussion

• Selective and macro editing and administrative
  data
• Potentials (use of admin data as auxiliary
  information for error modeling/error
  detection/error correction)
• Direct use of admin data for estimation purposes
  feasibility, costs, role of selective / macro
  editing
• Paper from Italy

13
Points for discussion

• Assessing the validity of a selective / macro
  editing strategy
• Which indicators, quality measures, evaluation
  approaches
• Inference and quality assessment under
  selectively edited data
• Countries experiences
• Paper from Italy, Sweden (Laitila-Norberg)

14
Points for discussion

• New, innovative approaches in selective / macro
  editing strategy
• Editing of unstructured data by using text
  analyses tools
• Using R-indicators for prioritization of
  follow-up
• Countries experiences
• Paper from United States , France (Merly-Alpa)