[FR?] Layered CVs grid: 10 time-series CV (window) * 10 spatial CVs = 100 CVs

[PathosEthosLogos]

First, thank you for this library. This library has been tremendous help.

https://github.com/tidymodels/rsample/issues/64#issue-362961204

The paper cited in the link above discusses blocked CVs.

In the latest release (0,2), it seems that layered CVs is the direction.

https://rsample.tidymodels.org/reference/slide-resampling.html

I would like to discuss the plausibility/practicality of a request for a two-layer blocked resampling method, where it combines (for example) the functions of rsample::sliding_window() and spatialsample::spatial_clustering_cv(). If sliding_window() creates 10 CVs and spatial_clustering_cv() creates 10 CVs, it would create a total of 100 CVs.

Please let me know what you think!

[mikemahoney218]

I believe this is possible via rsample::nested_cv(), for instance (using the attached ex.csv):

library(rsample)
library(spatialsample)

example <- read.csv("https://github.com/tidymodels/spatialsample/files/8984843/ex.csv")

x <- example |> 
  sliding_window(lookback = 1000)
pretty(x)
#> [1] "Sliding window resampling"

y <- example |> 
  spatial_clustering_cv(coords = c(x, y), v = 5)
y
#> #  5-fold spatial cross-validation 
#> # A tibble: 5 × 2
#>   splits              id   
#>   <list>              <chr>
#> 1 <split [8123/1877]> Fold1
#> 2 <split [8008/1992]> Fold2
#> 3 <split [8085/1915]> Fold3
#> 4 <split [7729/2271]> Fold4
#> 5 <split [8055/1945]> Fold5

xy <- example |> 
  nested_cv(
    outside = spatial_clustering_cv(coords = c(x, y), v = 5),
    inside = sliding_window(lookback = 1000)
  )

pretty(xy)
#> [1] "Nested resampling:"                     
#> [2] " outer: 5-fold spatial cross-validation"
#> [3] " inner: Sliding window resampling"

^{Created on 2022-06-25 by the reprex package (v2.0.1)}

Is this what you're thinking of?

ex.csv

[mikemahoney218]

I'm going to go ahead and close this out! If it turns out I missed something, please feel free to open a new issue with more information. Thank you so much for helping out!

[github-actions[bot]]

This issue has been automatically locked. If you believe you have found a related problem, please file a new issue (with a reprex: https://reprex.tidyverse.org) and link to this issue.