enEnglish
CROS

ST2_2 Overlapping numerical variables with a benchmark

Document date: 
Monday, 28 August, 2017
Language: 
English

ISTAT has also examined basic data configuration 2 for the situation where there are overlapping units as well as overlapping variables in the data sources to be combined and one of the data sources (typically a survey measure) is considered as error free and the administrative measures are merely used as auxiliary information.
Thus, differently from the situation in Istat Suitability test n.1, the error free variable (X) is modelled as a response variable and all the other measures (Y) are considered covariates. This supervised approach can be adopted in a model based inference as well as in a design based inference. In the latter case, the covariates can be used to specify a working model (model assisted approach). The choice of the methodological approach depends both on the informative content and the quality of the available data sources.