On Simulation about MPPS Sampling Method by Using the Census Data

摘要

MPPS sampling refers to the multivariate probability proportional to size sampling, which is proposed in 1990's. In recent years, the relevant survey agency in China conducted a pilot survey aiming on multi-purpose based on MPPS sample design which was collaborated with the National Agricultural Statistics Service (NASS) under the United States Department of Agriculture (USDA). Until now the MPPS sampling method is not widely applied in China. This paper gives a brief review of MPPS sampling method including its basic estimation procedure, and demonstrates a simulation results by using the real agricultural census data. In our simulation, we compared the results from the systematic sampling and Poisson sampling, and illustrated the design effects by using the census data. In addition, we simulated the results from the Poisson sampling with the permanent random number, and through our simulation we would like to make a clearance to one of its misuse by demonstrating the persuasive results.