Abstract
In this paper, we propose a new estimator for creating expansion factors for survey plots in the US Forest Service (USFS) Forest Inventory and Analysis program. This estimator was previously used in the GIS literature, where it was called Penalized Maximum Entropy Dasymetric Modeling. We show here that the method is a regularized version of the raking estimator widely used in sample surveys. The regularized raking method differs from other predictive modeling methods for integrating survey and ancillary data, in that it produces a single set of expansion factors that can have a general purpose which can be used to produce small-area estimates and wall-to-wall maps of any plot characteristic. This method also differs from other more widely used survey techniques, such as GREG estimation, in that it is guaranteed to produce positive expansion factors. Here, we extend the previous method to include cross-validation, and provide a comparison to expansion factors between the regularized raking and ridge GREG survey calibration.
Highlights
Surveys for inventories of forests, such as the United States Department of Agriculture ForestService (USFS) Forest Inventory and Analysis (FIA) program, are typically designed to provide reliable estimates of characteristics over large spatial units, such as states
We present a method for producing expansion factors more suitable for creating small-area estimates and wall-to-wall maps of survey characteristics
For the purpose of this article, the relevant factor is that these models generate predictions and standard errors for forest characteristics measured at FIA plots
Summary
Service (USFS) Forest Inventory and Analysis (FIA) program, are typically designed to provide reliable estimates of characteristics over large spatial units, such as states. The two methods differ statistically, the common theme is that they are both highly tailored to produce reliable estimates for a single response variable. While these approaches can Forests 2019, 10, 1045; doi:10.3390/f10111045 www.mdpi.com/journal/forests. We will briefly introduce the raking estimator, but will describe a different survey weighting approach—Generalized Regression (GREG)—that has been more fertile ground for recent innovation than raking. Let { Jj } be another set of areas, large or small, and possibly overlapping, for which estimates are desired, such as counties, the entire study area, management areas, or the individual pixels.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
