Comparisons between conditional Poisson sampling and Pareto πps sampling designs

Conditional Poisson sampling (CPS) and Pareto pi ps sampling (PPS) design belong to a class of sampling schemes which have inclusion probabilities proportional to given size measures (pi ps). Algorithms were introduced to calculate first- and second-order exact inclusion probabilities for both schemes. In this paper using those algorithms, both methods are compared by calculating Horvitz-Thompson estimators of populations totals and computing their biases and variances. They are also illustrated by examples, and the results in these are compared with known asymptotics. (C) 2000 Elsevier Science B.V. All rights reserved. MSG: primary 62D05; secondary 65U05.