Affiliations: [a] Brazilian Network Information Center (NIC.br), SP, Brazil | [b] National School of Statistical Sciences, RJ, Brazil
Corresponding author: Isabela Bertolini Coelho, Brazilian Network Information Center (NIC.br), Avenida das Nações Unidas, 11541 – 7 th floor, SP 04578-000, Brazil, E-mail: [email protected].
Abstract: Traditional surveys face increasing challenges due to rising non-response rates and the diminishing resources available to survey organizations. A recently proposed solution involves the combination of non-probability sample surveys (often cheaper) with probability sample surveys (more expensive), using the latter as a reference to weight the former. Considering a special case in which a single survey was designed and carried out by simultaneously using the two sampling approaches within a single field operation, this paper compared the use of quasi-randomization and sample matching methods to assign weights to the non-probability part of the sample. The quasi-randomization method provided the closest point estimates and smaller standard errors (on average) when compared to the benchmark estimates.