期刊名称：Documents de Travail du Centre d'Economie de la Sorbonne
印刷版ISSN：1955-611X
出版年度：2017
出版社：Centre d'Economie de la Sorbonne
摘要：Performing a statistical match to combine two surveys made over the samepopulation by traditional methods is shown to give biased estimates and varianceof the imputed values. A method proposed by Rubin (1986) allows imputingan unobserved variable using observations in another dataset by taking intoaccount the partial correlation between the variables that are jointly unobservedfor any unit. We use a dataset where households report their expenditures andtime-uses to show that fusioning expenditure and time-use surveys by Rubin’sprocedure allows to recover the true distribution of the missing variables andto yield minimally biased estimates.
关键词：Data combination; Data Fusion; Missing data; Statistical Matching; Time-use
