Model-Assisted Survey Estimation (mase)

Having precise estimates of our forest characteristics is important if we want to assess the status of our forests, detect change, or monitor trends. New statistical estimators enable us to improve precision by merging forest inventory data with data from a variety of remote sensing instruments but often pose computational challenges. This new tutorial and R software package, known as mase (model-assisted survey estimation) makes both old and new survey estimation tools easily accessible.

Purpose

Focusing on the broad class of model-assisted estimators under the umbrella of generalized regression estimators, we provide a tutorial that steps the reader through 7 estimators including Horvitz-Thompson, ratio, post-stratification, regression, lasso, ridge, and elastic net. Using forest inventory data from Daggett county in Utah as an example, we illustrate how to construct, as well as the relative performance of, these estimators. Each estimator is made readily accessible through the new R package, mase, available on the Comprehensive R Archival Network. We provide guidelines in the form of a decision tree on when to use which estimator in forest inventory applications.  

Citation

McConville, K. G.G. Moisen, T.S. Frescino. [In review.] A tutorial in model-assisted estimation with application to forest inventory. Canadian Journal of Forest Research.

McConville, K., B. Tang, G. Zhu, S. Cheung, and S. Li. 2017. mase: Model-Assisted Survey Estimation. R package version 0.1.1 https://github.com/Swarthmore-Statistics/mase.