Speaker: Dr. Emily Berg

Iowa State University, Department of Statistics

Methodology to Address Challenges in Small Area Estimation for Erosion 

Challenges in complex surveys motivate the use of model-based methods instead of standard design-based procedures. For instance, a remedy for non-response is to replace missing data with imputed values generated under a model. We briefly overview our recent work in model-based imputation. We focus primarily on small area estimation, the problem of constructing estimates for domains with small sample sizes. Small area estimates are of interest to the Natural Resources Conservation Service (NRCS) of the United States Department of Agriculture (USDA). NRCS monitors resource conditions on non-federal US land through the National Resources Inventory (NRI) and the Conservation Effects Assessment Project (CEAP). These probability-based surveys collect a rich set of characteristics related to landuse and erosion. We focus on variables that quantify soil and water losses from crop fields. The broad problem of small area estimation for erosion in the context of NRI and CEAP generates a variety of statistical issues. Distributions are zero-inflated. Quantiles are parameters of interest. The sample design is complex. Multivariate response variables have both discrete and continuous components. We develop procedures to address challenges that arise from the problem of small area estimation for measures erosion collected in the NRI and CEAP surveys.