Skip to contents

Prepare spatial-block cross-validation folds for spatial analysis

Source: R/mod_cv_prepare.R
mod_CV_prepare.Rd

This function assign modelling input data into spatial-block cross-validation folds using three strategies (see below) using blockCV::cv_spatial. The function is planned to be used inside the mod_prepare_hpc function.

Usage

mod_cv_prepare(
  input_data = NULL,
  env_file = ".env",
  x_vars = NULL,
  cv_n_folds = 4L,
  cv_n_grids = 20L,
  cv_n_rows = 2L,
  cv_n_columns = 2L,
  cv_sac = FALSE,
  out_path = NULL
)

Arguments

input_data

data.frame. A data frame or tibble containing the input dataset. This data frame should include two columns for x and y coordinates as long as other columns matching the names of predictors listed in x_vars argument. This argument is mandatory and can not be empty.

env_file

Character. Path to the environment file containing paths to data sources. Defaults to .env.

x_vars

Character vector. Variables to be used in the model. This argument is mandatory and can not be empty.

cv_n_folds

Integer. Number of cross-validation folds. Default: 4L.

cv_n_grids

Integer. Number of grid cells in both directions used in the cv_dist cross-validation strategy (see below). Default: 20L.

cv_n_rows, cv_n_columns

Integer. Number of rows and columns used in the cv_large cross-validation strategy (see below), in which the study area is divided into large blocks given the provided cv_n_rows and cv_n_columns values. Both default to 2L which means to split the study area into four large blocks at the median latitude and longitude.

cv_sac

Logical. Whether to use the spatial autocorrelation to determine the block size. Defaults to FALSE,

out_path

Character. Path for directory to save the cross-validation results. This argument is mandatory and can not be empty.

Value

The function returns a modified version of the input dataset with additional numeric columns (integer) indicating the cross-validation strategy used.

Note

The function uses the following cross-validation strategies:

  • cv_dist in which the size of spatial cross-validation blocks is determined by the cv_n_grids argument. The default cv_n_grids value is 20L, which means blocks of 20×20 grid cell each.

  • cv_large which splits the study area into large blocks, as determined by the cv_n_rows and cv_n_columns arguments. if cv_n_rows = cv_n_columns = 2L (default), four large blocks will be used, split the study area at the median coordinates.

  • cv_sac in which the size of the blocks is determined by the median spatial autocorrelation range in the predictor data (estimated using blockCV::cv_spatial_autocor). This requires the availability of the automap R package. This strategy is currently skipped by default.

Author

Ahmed El-Gabbas