Title: The ELFE survey and its complex survey design
Authors: Anne Ruiz-Gazen - Toulouse School of Economics (France) [presenting]
Abstract: The 2011 French Longitudinal Survey on Childhood ELFE comprises more than 18,000 children selected on the basis of their place and date of birth. A sample of 320 maternity units and a sample of 25 days divided in four time periods have been selected. The sample is made of babies born at the sampled locations and on the sampled days. The objective of this large survey is to analyze the physical and psychological health of children together with their living and environmental conditions. In order to derive reliable confidence intervals for finite population parameters, the ELFE sampling design has to be taken into account. This design consists in drawing samples from a two-dimension population, independently in each dimension, and is called the Cross-Classified Sampling design. A general theory of estimation is derived for this design. Generally, the cross-classified design results in a loss of efficiency as compared to the usual two-stage design. Even in the case of simple random sampling designs without replacement in each dimension, the usual variance estimators may be negative. Non-negative simplified variance estimators are introduced and compared through a simulation study. Finally, an application to the ELFE data is detailed.