Abstract:
|
The USDA's National Agricultural Statistics Service (NASS) is exploring sampling approaches that allow for coordination of multiple samples drawn within a year across the population in an effort to control the respondent burden. Most of these sampling techniques, which are both design-based and model-based approaches, utilize permanent random numbers (PRN) for the purpose of achieving the amount of desired overlap within a survey or between different surveys and, therefore, help reduce the respondent burden. However, little discussion comparing design-based and model-based approaches or examining a combination of these two approaches appears in the literature. Using a simulation study, we investigate different sampling strategies (a combination of sampling design and estimator) that utilize both design-based and model-based inferences. Our simulations are based on data from several USDA surveys. Results are presented.
|