ldmppr (version 1.1.0)

check_model_fit: Check the fit of estimated self-correcting model on the reference point pattern dataset

Description

Allows the user to perform global envelope tests for the nonparametric L, F, G, J, E, and V summary functions from the spatstat package. These tests serve as a goodness of fit measure for the estimated model relative to the reference dataset of interest.

Usage

check_model_fit(
  reference_data,
  t_min = 0,
  t_max = 1,
  sc_params = NULL,
  anchor_point = NULL,
  raster_list = NULL,
  scaled_rasters = FALSE,
  mark_model = NULL,
  xy_bounds = NULL,
  include_comp_inds = FALSE,
  thinning = TRUE,
  correction = "none",
  competition_radius = 15,
  n_sim = 2500,
  save_sims = TRUE,
  verbose = TRUE,
  seed = 0
)

Value

a list containing a combined global envelope test, individual global envelope tests for the L, F, G, J, E, and V functions, and simulated metric values (if specified).

Arguments

reference_data

a ppp object for the reference dataset.

t_min

minimum value for time.

t_max

maximum value for time.

sc_params

vector of parameter values corresponding to (alpha_1, beta_1, gamma_1, alpha_2, beta_2, alpha_3, beta_3, gamma_3).

anchor_point

vector of (x,y) coordinates of point to condition on.

raster_list

a list of raster objects.

scaled_rasters

`TRUE` or `FALSE` indicating whether the rasters have been scaled.

mark_model

a model object (typically from train_mark_model).

xy_bounds

a vector of domain bounds (2 for x, 2 for y).

include_comp_inds

`TRUE` or `FALSE` indicating whether to generate and use competition indices as covariates.

thinning

`TRUE` or `FALSE` indicating whether to use the thinned or unthinned simulated values.

correction

type of correction to apply ("none" or "toroidal").

competition_radius

distance for competition radius if include_comp_inds is `TRUE`.

n_sim

number of simulated datasets to generate.

save_sims

`TRUE` or `FALSE` indicating whether to save and return the simulated datasets.

verbose

`TRUE` or `FALSE` indicating whether to show progress of model checking.

seed

an integer value to set the seed for reproducibility.

Details

This function relies on the spatstat package for the calculation of the point pattern metrics and the GET package for the global envelope tests. The L, F, G, J, E, and V functions are a collection of non-parametric summary statistics that describe the spatial distribution of points and marks in a point pattern. See the documentation for [spatstat.explore::Lest()], [spatstat.explore::Fest()], [spatstat.explore::Gest()], [spatstat.explore::Jest()], [spatstat.explore::Emark()], and [spatstat.explore::Vmark()] for more information. Also, see the [GET::global_envelope_test()] function for more information on the global envelope tests.

References

Baddeley, A., Rubak, E., & Turner, R. (2015). *Spatial Point Patterns: Methodology and Applications with R*. Chapman and Hall/CRC Press, London. ISBN 9781482210200. Available at: https://www.routledge.com/Spatial-Point-Patterns-Methodology-and-Applications-with-R/Baddeley-Rubak-Turner/p/book/9781482210200.

Myllymäki, M., & Mrkvička, T. (2023). GET: Global envelopes in R. arXiv:1911.06583 [stat.ME]. tools:::Rd_expr_doi("10.48550/arXiv.1911.06583").

Examples

# Note: The example below is provided for illustrative purposes and may take some time to run.
# \donttest{
# Load the small example data
data(small_example_data)

# Load the example mark model that previously was trained on the small example data
file_path <- system.file("extdata", "example_mark_model.rds", package = "ldmppr")
mark_model <- load_mark_model(file_path)

# Load the raster files
raster_paths <- list.files(system.file("extdata", package = "ldmppr"),
                           pattern = "\\.tif$", full.names = TRUE)
raster_paths <- raster_paths[!grepl("_med\\.tif$", raster_paths)]
rasters <- lapply(raster_paths, terra::rast)

# Scale the rasters
scaled_raster_list <- scale_rasters(rasters)

# Generate the reference pattern
reference_data <- generate_mpp(
  locations = small_example_data[, c("x", "y")],
  marks = small_example_data$size,
  xy_bounds = c(0, 25, 0, 25)
)

# Define an anchor point
M_n <- c(small_example_data[1, c("x", "y")])

# Specify the estimated parameters of the self-correcting process
# Note: These would generally be estimated using estimate_process_parameters.
# These values are estimates from the small_example_data generating script.
estimated_parameters <- c(
  0.05167978, 8.20702166, 0.02199940, 2.63236890,
  1.82729512, 0.65330061, 0.86666748, 0.04681878
)

# Check the model fit
example_model_fit <- check_model_fit(
  reference_data = reference_data,
  t_min = 0,
  t_max = 1,
  sc_params = estimated_parameters,
  anchor_point = M_n,
  raster_list = scaled_raster_list,
  scaled_rasters = TRUE,
  mark_model = mark_model,
  xy_bounds = c(0, 25, 0, 25),
  include_comp_inds = TRUE,
  thinning = TRUE,
  correction = "none",
  competition_radius = 10,
  n_sim = 100,
  save_sims = FALSE,
  verbose = TRUE,
  seed = 90210
)

plot(example_model_fit, which = 'combined')
# }