~~l United States
SRB Research Report
An Evaluation of Truncation
Estimators for Improving
State Estimates of Total
AN EVALUATION OF TRUNCATION ESTIMATORS FOR IMPROVING STATE
ESTIMATES OF TOTAL HOGS, by Susan Hicks, Matt Fetter, and Susan Cowles, Sampling
and Estimation Research Section, Survey Research Division, National Agricultural Statistics
Service, United States Department of Agriculture, Washington, D.C. 20250-2000, April 1995,
Report No. SRB-95-02.
The National Agricultural Statistics Service (NASS) consistently strives to produce more reliable
estimates of hogs at both the State and National levels through its Quarterly Agricultural Surveys
(QAS). One contributor to high variances at the State level are outliers in the survey data.
Outliers are defined as extremely large expanded observations. Outliers can be caused by large
weights, large reported values of hogs, or a combination of both. In this report, we evaluate the
efficiency of a class of estimators, called weight truncation estimators, for improving the
reliability of the State level estimates of total hogs in the presence of outliers. The estimators
truncate the weights of the large expanded observations so that the expanded value does not
exceed a predetermined cutoff. Two types of truncation estimators are evaluated: data-driven
estimators and fixed cutoff estimators. The fixed cutoff estimators hold the most promise for
improving the State level estimates both in terms of ease of application and efficiency of the
Robust Estimation; Truncation-type Estimators; Winsorization; Minimum Estimated MSE
Trimming; Iterative Fence Method.
This paper was prepared for limited distribution to the research community outside the
U.S. Department of Agriculture. The views expressed herein are not necessarily those
of the USDA or NASS.