Skip to contents

Splits the sites into k folds, either as spatial blocks, clusters of neighbouring sites found by k-means on the coordinates, or at random. With spatial autocorrelation, sites left out at random have close neighbours among the sites kept, so random folds can make a model look better at predicting new sites than it is; spatial blocks hold out whole regions (Roberts et al. 2017).

Usage

spatial_folds(coords, k = 5, method = c("blocks", "random"), seed = 1)

Arguments

coords

Matrix or data frame with one row per site and the coordinates in columns, such as landscape$coords or the coords of an mc_data().

k

Number of folds.

method

"blocks" for spatial blocks or "random".

seed

Integer that sets the seed.

Value

Integer vector with the fold of each site.

References

Roberts, D. R., Bahn, V., Ciuti, S., Boyce, M. S., Elith, J., Guillera-Arroita, G., Hauenstein, S., Lahoz-Monfort, J. J., Schröder, B., Thuiller, W., Warton, D. I., Wintle, B. A., Hartig, F., & Dormann, C. F. (2017). Cross-validation strategies for data with temporal, spatial, hierarchical, or phylogenetic structure. Ecography, 40, 913–929. doi:10.1111/ecog.02881

Examples

grid <- landscape_grid(side = 6)
table(spatial_folds(grid$coords, k = 4))
#> 
#> 1 2 3 4 
#> 9 9 9 9