Analysing correspondence between administrative and survey data.

Autor: van Delden, Arnout, Pannekoek, Jeroen, Banning, Reinder, de Boer, Arjen
Předmět:
Zdroj: Statistical Journal of the IAOS; 2016, Vol. 32 Issue 4, p569-584, 16p
Abstrakt: For National Statistical Institutes, administrative data sources are an attractive alternative to survey sampling to produce official statistics. The question may arise whether a candidate administrative data source can replace existing survey sampling, for a range of publication domains. We combined an analysis of conceptual differences with indicators from a robust linear regression analysis to assign the domains to four groups: a Control group without conceptual differences, an Accept group with conceptual differences but only small numerical differences, an Adjust group with conceptual differences and substantial, systematic, numerical differences and a Reject group with conceptual differences and substantial, non-systematic, numerical differences. We applied our approach to Value Added Tax data at Statistics Netherlands from which quarterly and yearly turnover statistics are derived. Using the range of values for the indicators of the Control group, we derived thresholds to sort the remaining domains into the groups Accept, Adjust and Reject. Our approach can be used as a first means to sort the domains, but additional analyses and sometimes additional collection of data is needed when conceptual and numerical results are not in line with each other. [ABSTRACT FROM AUTHOR]
Databáze: Complementary Index