Dynamic K-Composite Estimator for an Arbitrary Rotation Scheme
Dynamic K-Composite Estimator for an Arbitrary Rotation Scheme

Summary/Abstract: Classical K-composite estimator was proposed in Hansen et al. (1955). Its optimality properties were developed in Rao and Graham (1964). This estimator gives an alternative solution to quasi-optimal estimation under rotation sampling when it is allowed that units leave the sample for several occasions and then come back. Such situations happen frequently in real surveys and are not covered by the recursive optimal estimator introduced by Patterson (1955). However the K-composite estimator suffers from certain disadvantages. It is designed for a stable situation in the sense that its basic parameter is kept constant on all occasions. Additionally it is restricted only to a certain family of rotation designs. Here we propose a dynamic version of the K-composite estimator (DK-composite estimator) without any restrictions on the rotation pattern. Mathematically, the algorithm, we develop, is much simpler than the one for the classical K-composite estimator with optimal weights. Moreover, it is precise, in the sense that it does not use any approximate or asymptotic approach (opposed to the method used in Rao and Graham (1964) for computing optimal weights).

• Details
• Contents