Journal of Official Statistics, Vol.13, No.4, 1997. pp. 367–383
Developing an Estimation Strategy for a Pesticide Data Program
Phillip S. Kott and D. Andrew Carr
Abstract:The Agricultural Marketing Service's Pesticide Data Program (PDP) is a cooperative effort of U.S. Department of Agriculture (USDA) and several state agencies. The ultimate purpose of the program is to make scientific statements about the distribution of certain pesticide residues in particular products (mostly fresh fruits and vegetables) consumed by the U.S. public. Developing a statistically defensible estimation strategy for the PDP required overcoming a number of thorny problems. Chief among them was the non-random nature of the “sample” of participating states. Also of concern was the level-of-detection/level-of-quantification issue: not all potential levels of pesticide residue can be detected by a given lab; moreover, certain detectable levels are not quantifiable. A graphical method was developed to display parameter estimates (means and percentiles) in light of the detection/quantification problem. Included on the graphs (as an option) are fairly robust, model-based estimates of confidence intervals.
Keywords:Target population; inferential population; level of detection; level of quantification; percentile; distribution.
Copyright © Statistics Sweden 1996-2018. Open AccessISSN 0282-423XCreated and Maintained by OKS Group