PLACE:  319 Snedecor

SPEAKER:
Jean Opsomer
Iowa State University

TITLE:
Local Polynomial Regression Estimation In Complex Survey

ABSTRACT:

In survey statistics, the study objective is often to estimate the finite population total or mean for a variable of interest. Auxiliary information can be used to improve upon classical design-based estimators by constructing model-assisted estimators, which have traditionally been derived from parametric models.  We introduce the local polynomial regression estimator, a nonparametric model-assisted estimator, and describe its theoretical properties. The approach generalizes readily to more complex designs such as multi-phase and multi-stage designs.  In this talk, we extend the estimator to the case of two-stage (cluster) sampling.  The behavior of the method is illustrated through simulation experiments and an example of erosion estimation using 1995 data from the National Resources Inventory.
 

COFFEE: 3:45 p.m., 104 Snedecor