Complex Sample Survey Estimation in Static State-Space

摘要

Increased use of remotely sensed data is a key strategy adopted by the Forest Inventory and Analysis Program. However, multiple sensor technologies require complex sampling units and sampling designs. The Recursive Restriction Estimator (RRE) accommodates this complexity. It is a design-consistent Empirical Best Linear Unbiased Prediction for the state-vector, which contains all suffi cient statistics for the sampled population. RRE reduces a complex estimator into a sequence of simpler estimators. Also included are model-based pseudo-estimators and multivariate Taylor series approximations for covariance matrices. Together, these provide a unifi ed approach to detailed estimation in large, complex sample surveys.