Online Program

Return to main conference page
Friday, May 18
Computational Statistics
Survey Science
Fri, May 18, 10:30 AM - 12:00 PM
Lake Fairfax B

Survey Estimation with Elastic Net Regression: Combining Data Sources to Improve Estimator Efficiency (304310)

Jay Breidt, Colorado State University 
Tracey Frescino, U.S. Department of Agriculture Forest Service  
Thomas Lee, University of California, Davis 
*Kelly Sue McConville, Swarthmore College 
Gretchen Moisen, U.S. Department of Agriculture Forest Service 

Keywords: survey sampling, model-assisted estimation, elastic net regression, lasso

Often in natural resource surveys where the goal is to estimate population quantities, auxiliary data can supplement the survey data and improve estimator efficiency. One common technique is to build a model-assisted regression estimator where a linear model is assumed for the relationship between the survey variable and auxiliary variables. We propose estimating the coefficients of the model using elastic net regression to control for extraneous variables and multicollinearity. Using data from the Forest Inventory and Analysis Program, we compare the model-assisted estimators based on regularized regression to other survey estimators.