Project details :
Data Quality in Complex Surveys within the New European Information Society
03/2001 to 05/2004
1.33 million euro
2.16 million euro
Dr. Ian Perry (email@example.com)
Information Society Technologies - 5th FWP (Fifth Framework Programme) of the EC
Project Main Goals:
The European statistics faces the difficult task of creating a harmonised and reliable socioeconomic database for the New Economy in a united Europe. The definitions used in the member states of the EU need to be standardised and the quality of the data gained from complex surveys need to be made more homogeneous. The core of the problem is to obtain practical and usable methods for variance estimation in complex multi-purpose sampling, which enable the user to effectively and reliably estimate variances with comparable standards for the relevant national surveys by choosing from a list of criteria. This leads to a harmonised and standardised European quality management system in statistics. To fulfil this task, all relevant variance estimation methods currently available need to be analysed, classified, evaluated and improved. This will be achieved firstly by theoretical research secondly by a Monte-Carlo study with simulated realistic universes of the relevant national surveys
Errors in data may have unfortunate consequences in economic and social applications if they remain unknown. Therefore the aim of DACSEIS is to advance the provision of reliable information on data quality to accompany statistics databases. Best practice recommendations will be prepared for users as well as a discussion of the possibilities for harmonising sampling methods.
A main target of DACSEIS is the standardisation and harmonisation of variance estimation methods used to calculate sampling errors, with special emphasis on different national surveys within the European statistical system and their international comparison. A catalogue of applicable instructions and criteria is to be prepared allowing the user to choose among effective procedures of variance estimation for complex sample designs. Simulated complex universes, reflecting the relevant national surveys and their characteristics, are generated to analyse the influence of special designs and arrangements of subuniverses on the precision of the variance estimation methods. Standard software packages for survey sampling are examined and evaluated with special emphasis on the implementation of the variance estimation methods for complex surveys mentioned above or to their extensibility. Results as well as estimation methods will be put at the disposal of official, non-official, and other interested users with respect to high quality standards in statistics. This includes the delivery of a best practice manual as well as the source and pseudo codes of all relevant estimation procedures.
- Substantial compilation, classification and evaluation of all relevant methods for variance estimation in complex survey designs
- Development of a simulative generation of relevant socio-economic universes with the emphasis on a realistic Monte-Carlo-study of the methods
- Monte-Carlo-simulations for the evaluation of the procedures and their enhancement;
- Theoretical research with the improvement of procedures
- Catalogue of user-oriented instructions and criteria for variance estimation methods in practical use
The research project DACSEIS provides an European quality-management-system for socio-economic data in the national statistical sectors obtained by complex survey sampling. The core of the problem of quality measurement for survey data is the development of variance estimation methodology for complex multi-purpose sampling schemes. Complex samples are currently used in all national official and non-official statistic institutes to guarantee a broad socio-economic database.
The major aim of the approach is the consistent classification, methodological unification, evaluation, and improvement of existing methods for variance estimation in complex survey samples. The methods to be investigated include resampling methods, methods for unequal probability designs, methods for combining register and survey data, raking adjustment methods, and variance estimation methods for change. Further estimation procedures that focus on special topics, e.g. small area estimation, are examined.
All relevant methods are to be inspected, analysed, and evaluated with respect to their applicability to complex surveys. This will be achieved by simulated but realistic complex universes as a basis for a detailed Monte-Carlo simulation study of the surveying methods and their variance estimation. Computer codes for all relevant variance estimation methods will be developed. Additionally, standard software packages for survey sampling are examined and evaluated regarding the variance estimation methods considered. The dissemination of the DACSEIS research will take place by traditional publication procedures, a final international conference, and a public WWW-server.