EcoSta 2022: Start Registration
View Submission - EcoSta2022
A0939
Title: Improving finite population inference by data integration Authors:  Anne Ruiz-Gazen - Toulouse School of Economics (France) [presenting]
Estelle Medous - University of Toulouse 1 (France)
Camelia Goga - CNRS-LMB (France)
Jean-Francois Beaumont - Statistics Canada (Canada)
Alain Dessertaine - La Poste (France)
Pauline Puech - La Poste (France)
Abstract: Combining survey sample data and big databases is an important current challenge in finite population inference. While survey sample data are obtained through a probability sampling design, big data consist usually of non-probability samples. Many well-known unbiased or approximately unbiased methods exist for estimating finite population parameters from a probability sample. Inference from a non-probability sample is, however, often subject to selection bias. Recently, a data integration approach has been proposed that allows handling the selection bias of non-probability samples by incorporating a probability sample. We propose to revisit their approach and study in detail the gain in terms of efficiency of the estimators based on a probability sample when taking into account non-probability samples.