blockCV: blockCV: A package for generating spatially or...
In adamlilith/blockCV: Spatial and environmental blocking for k-fold cross-validation
Description
Author(s)
References
See Also
Description
Simple random selection of testing and training folds in structured environment leads to underestimation of error in the evaluation of spatial
predictions and may result in inappropriate model selection (Telford and Birks, 2009; Roberts et al., 2017). The use of spatial and
environmental blocks to separate test and train sets has been suggested as a good strategy for realistic error estimation in datasets
with dependence structures, and more generally as a robust method for estimating predictive performance of models used to predict mapped
distributions (Roberts et al., 2017).
Package blockCV provides functions to separate train and test sets using buffers, spatial and environmental
blocks. It provides several options for how those blocks are constructed.
It also has a function that applies geostatistical techniques to investigate the existing
level of spatial autocorrelation in the covariates to inform the choice of a suitable distance band by which to separate the data sets.
In addition, some visualization tools are provided to help the user choose the block size and explore generated folds. The package has been
written with species distribution modelling in mind, and the functions allow for a number of common scenarios (including presence-absence
and presence-background species data, rare and common species, raster data for predictor variables).
Author(s)
Roozbeh Valavi, Jane Elith, José Lahoz-Monfort and Gurutzeta Guillera-Arroita
References
Roberts et al., 2017. Cross-validation strategies for data with temporal, spatial, hierarchical,
or phylogenetic structure. Ecography. 40: 913-929.
Telford, R.J., Birks, H.J.B., 2009. Evaluation of transfer functions in spatially structured environments. Quat. Sci. Rev. 28, 1309–1316.
See Also
spatialBlock, buffering and envBlock for blocking strategies.
adamlilith/blockCV documentation built on May 25, 2019, 12:41 a.m.
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Description Author(s) References See Also
Description
Simple random selection of testing and training folds in structured environment leads to underestimation of error in the evaluation of spatial predictions and may result in inappropriate model selection (Telford and Birks, 2009; Roberts et al., 2017). The use of spatial and environmental blocks to separate test and train sets has been suggested as a good strategy for realistic error estimation in datasets with dependence structures, and more generally as a robust method for estimating predictive performance of models used to predict mapped distributions (Roberts et al., 2017). Package blockCV provides functions to separate train and test sets using buffers, spatial and environmental blocks. It provides several options for how those blocks are constructed. It also has a function that applies geostatistical techniques to investigate the existing level of spatial autocorrelation in the covariates to inform the choice of a suitable distance band by which to separate the data sets. In addition, some visualization tools are provided to help the user choose the block size and explore generated folds. The package has been written with species distribution modelling in mind, and the functions allow for a number of common scenarios (including presence-absence and presence-background species data, rare and common species, raster data for predictor variables).
Author(s)
Roozbeh Valavi, Jane Elith, José Lahoz-Monfort and Gurutzeta Guillera-Arroita
References
Roberts et al., 2017. Cross-validation strategies for data with temporal, spatial, hierarchical, or phylogenetic structure. Ecography. 40: 913-929.
Telford, R.J., Birks, H.J.B., 2009. Evaluation of transfer functions in spatially structured environments. Quat. Sci. Rev. 28, 1309–1316.
See Also
spatialBlock, buffering and envBlock for blocking strategies.
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Embedding an R snippet on your website
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.