Bias correction of bounded location errors in presence‐only data
| Autor: | Brian M. Brost, Mevin B. Hooten, Trevor J. Hefley |
|---|---|
| Rok vydání: | 2017 |
| Předmět: |
0106 biological sciences
Location test Ecological Modeling Inference 010603 evolutionary biology 01 natural sciences Hierarchical database model 010104 statistics & probability Transformation (function) Statistics Poisson point process Errors-in-variables models 0101 mathematics Spatial analysis Ecology Evolution Behavior and Systematics Sign (mathematics) Mathematics |
| Zdroj: | Methods in Ecology and Evolution. 8:1566-1573 |
| ISSN: | 2041-210X |
| DOI: | 10.1111/2041-210x.12793 |
| Popis: | Summary Location error occurs when the true location is different than the reported location. Because habitat characteristics at the true location may be different than those at the reported location, ignoring location error may lead to unreliable inference concerning species–habitat relationships. We explain how a transformation known in the spatial statistics literature as a change of support (COS) can be used to correct for location errors when the true locations are points with unknown coordinates contained within arbitrary shaped polygons. We illustrate the flexibility of the COS by modelling the resource selection of Whooping Cranes (Grus americana) using citizen contributed records with locations that were reported with error. We also illustrate the COS with a simulation experiment. In our analysis of Whooping Crane resource selection, we found that location error can result in up to a five-fold change in coefficient estimates. Our simulation study shows that location error can result in coefficient estimates that have the wrong sign, but a COS can efficiently correct for the bias. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|
Plný text