Small-Area Estimation, Sampling Adjustment & Calibration

Objectifs

The first part (2 days) focuses on estimation problems on small domains. When the sub-populations of interest are small, the overlapping sample size is small and leads to highly unstable estimators. It is then necessary to mobilize all available auxiliary information, in order to propose estimation methods based on survey theory and a modeling approach.

The second part (2 days) covers adjustment techniques (also known as "calibration"). Once a sample has been drawn, the initial set of weights is modified at the margin to significantly reduce the sampling variance, while retaining the essential property of being virtually unbiased. This methodology can be applied in a variety of contexts, in particular when the theoretical conditions for standard implementation are not met. A half-day computerized calibration practice session (standard calibration) is planned.

Plan

Part 1: Estimation on small estates

– Direct estimation
– Indirect estimation with implicit modeling
– Indirect estimation with explicit modeling – case of quantitative variables
– Indirect estimation with explicit modeling – case of qualitative variables

Part 2: Rectification: reminders and additions
– Review of standard theory
– Application to domains
– Multilevel calibration
– Context of obsolete auxiliary information
– Context of uncertain auxiliary information
– Context of little or non-explicative auxiliary information
– Calculation and truncated functions: choice of bounds
– Practical work on computer (1/2 day)

Prérequis

Not indicated