Grouped cross-validation folds for community data

Samples from the same site (or transect, plot, ...) are not independent, so random cross-validation folds leak information and overstate model skill (Roberts et al. 2017). model_group_folds() creates folds that keep every row of a group together, using rsample::group_vfold_cv().

Usage

model_group_folds(data, group, v = NULL, ...)

Arguments

data

A data frame, typically one row per sampling unit.

group

<tidy-select> The blocking column, such as site.

v

Number of folds. Defaults to the number of groups (leave-one-group-out).

...

Passed to rsample::group_vfold_cv().

Value

An rset object for use with tune.

References

Roberts, D. R. et al. (2017). Cross-validation strategies for data with temporal, spatial, hierarchical, or phylogenetic structure. Ecography, 40, 913-929. doi:10.1111/ecog.02881

Examples

units <- div_hill(png_ants, q = 0, .by = c(season, transect))
model_group_folds(units, group = site)
#> # Group 5-fold cross-validation 
#> # A tibble: 5 x 2
#>   splits         id       
#>   <list>         <chr>    
#> 1 <split [16/4]> Resample1
#> 2 <split [16/4]> Resample2
#> 3 <split [16/4]> Resample3
#> 4 <split [16/4]> Resample4
#> 5 <split [16/4]> Resample5