Improving Discovery-Based Probability Models for the Shoshone National Forest, Wyoming
Author(s): Paul Burnett; Erik Otarola-Castillo; Lawrence Todd
Year: 2016
Summary
Site prediction models continue to contribute useful information to the management of archaeological resources. For example, since 2009 we have developed several probability models for the Shoshone National Forest. The first model was used to guide inventory of areas burned in wildland fires to rapidly appraise archaeologically sensitive areas. The model was overhauled in 2015 to cover the entire Shoshone National Forest. Until now, we have used stepwise logistic regression to identify environmental parameters significantly contributing to site presence/absence. This approach has been successful. However, opportunities for model improvement remain. First, the stepwise regression procedure can be affected by the happenstance of a particular random sample. In addition, estimates of model parameters can be heavily biased due to sample area coverage and its resulting spatial autocorrelation. We improve on this process by employing new statistical techniques guarding against pseudoreplication of variables by modeling and accounting for the observed autocorrelation in the site/nonsite sample. Secondly, rather than relying on stepwise selection of a variety of environmental parameters for inclusion in the model, we employ environmental parameters that we interpret as conditioning site presence/absence. We use the Akaike Information Criterion to identify the best model.
Cite this Record
Improving Discovery-Based Probability Models for the Shoshone National Forest, Wyoming. Paul Burnett, Erik Otarola-Castillo, Lawrence Todd. Presented at The 81st Annual Meeting of the Society for American Archaeology, Orlando, Florida. 2016 ( tDAR id: 405116)
This Resource is Part of the Following Collections
Keywords
General
Predictive Modeling
Geographic Keywords
North America - Plains
Spatial Coverage
min long: -113.95; min lat: 30.751 ; max long: -97.163; max lat: 48.865 ;