TECHNICAL NOTE DATA QUALITY
NON-SAMPLING ERROR
1 Non-sampling errors may arise as a result of errors in the reporting or processing of data. These errors can be introduced through inadequacies in the questionnaire, treatment of non-response, inaccurate reporting by data providers, errors in the application of survey procedures, incorrect recording of answers and errors in data capture and processing.
2 The extent to which non-sampling error affects the results is difficult to measure. Every effort is made to minimise non-sampling error by careful design and testing of the collection instrument, the use of efficient operating procedures and systems, and the use of appropriate methodologies.
Reliability of Statistics
3 When interpreting the statistics in this release, the reliability and comparability of the estimates may be affected by the following specific non-sampling errors:
- Many organisations provided estimates due to a lack of separately recorded data on R&D activity. This was most prevalent for government organisations without a specific research focus.
- Data were subjectively classified, by organisations, to Research fields, Socio-economic objectives and Type of activity at the time of reporting. Some organisations may have experienced difficulty in classifying their R&D projects. The ABS makes every effort to ensure correct and consistent interpretation and reporting of these data by applying consistent processing methodologies.
- The estimation method for R&D related overhead costs varied across organisations and reference periods.
REVISIONS
4 Revisions to previous cycle data occur on discovery of errors in previously reported data. Errors identified are typically a result of the specific non-sampling errors outlined in the Reliability of Statistics section above.
5 This release includes revised data for the 2004-05 reference period. Revisions were primarily the result of: provider reassessment of application of definitions and classifications; and differences in interpretation due to changes in personnel. The effect of revisions is most noticeable in component item data. Revisions have been applied where the impact on:
- R&D expenditure is equal to $5 million or more;
- Human resources devoted to R&D is equal to 25 PYE or more; or
- Published level data is of proportional significance.
6 In processing 2006-07 data, revisions were not applied to data in cycles prior to 2004-05. This must be taken into consideration when interpreting results, particularly when comparing estimates over time.
Print Page
Print All