Presentation is loading. Please wait.

Presentation is loading. Please wait.

Improving the efficiency of editing in ONS business surveys

Similar presentations


Presentation on theme: "Improving the efficiency of editing in ONS business surveys"— Presentation transcript:

1 Improving the efficiency of editing in ONS business surveys
Rachel Skentelbery Office for National Statistics, UK

2 Overview Introduction to Eden Project
Selective Editing Methodology – Short term surveys Annual Business Survey Overview of SELEKT Results Collaborations and next steps

3 Eden Project New editing strategy for the ONS following visit in 2007 from Australian Bureau of Statistics Aim to improve the efficiency of editing process for business surveys, based on Australian model Deliver efficiencies with little impact on quality Better balance between micro and macro editing Implementation of new selective editing for RSI and MBS

4 Eden Principles 1. Holistic Approach 2. Sound Methods 3. Maximise Impact of Resources 4. Process Quality 5. Informed Data 6. Continuous Quality Improvement

5 Selective editing methodology - STS
Targets potential errors which have a significant impact on key estimates Selective editing works by: assigning a score to each business; the score reflects the impact that editing the response would have on the estimates; businesses with a score above their domain threshold are validated; those businesses with a score below the threshold are not validated.

6 Quality Measures - ARB Absolute relative bias aims to control the residual bias left in the domain estimates after editing

7 Savings Savings measure the change in the number of units that will be manually micro edited

8 Results – Short Terms Surveys
The number of forms failing has reduced by approximately 20% on both RSI and MBS Quality – ARB below 1% for all domains Monitoring a number of ‘quality indicators’

9 Annual Business Survey (ABS)
Large business survey split by different sectors of the economy Many questions asked from each business Large number of edit rules applied to identify suspect values in returned data

10 Alternative approach needed for ABS
Current selective editing is unsuitable for ABS due to Large number of (key) survey variables Lack of sufficient predictors needed to calculate scores Collaboration between ONS, Statistics Sweden and Pedro Silva (University of Southampton) to investigate use of SELEKT

11 SELEKT Tool Generic software to perform Selective Editing
Developed at Stats Sweden by A. Nordberg and team Set of SAS macros Driven by large set of user-specified parameters Can deal with large, complex surveys like ABS

12 SELEKT Scoring Method Local scores defined for each combination of:
Domain of interest (d) Variable (j) Record / unit (i) Local score used in SELEKT considers three components: Potential impact Suspicion Importance Global score is obtained by aggregating local scores

13 Results from testing Production & Construction Sector Catering Sector
Total number of records: 4,164 Records flagged by original edits: 2,923 Catering Sector Total number of records: 784 Records flagged by original edits: 620 Global threshold Records flagged - SELEKT Savings % Average Global Bias (all) % Average Global Bias (key) % 0.0005 2259 22.7 0.7 0.3 Global threshold Records flagged - SELEKT Savings % Average Global Bias (all) % Average Global Bias (key) % 0.0001 483 22.1 0.8 0.2

14 Work in progress Completed feasibility study – looking to implement
Fine-tuning parameters for P&C and Catering Also testing Statistics Sweden LAB to help choose best performing parameters Set up parameter table for all other sectors Plans for use in ABS, 2012 dispatch


Download ppt "Improving the efficiency of editing in ONS business surveys"

Similar presentations


Ads by Google