PSTestRun: Perform the Monte Carlo test for preferential sampling

In joenomiddlename/PStestR: Monte Carlo Test for Preferential Sampling in Spatio-Temporal Data

Perform the Monte Carlo test for preferential sampling

Description

PSTestRun returns the empirical pointwise p-values of the test. Additionally, it may also return simultaneous empirical p-values from a rank envelope test with plots if asked.

This function is called after initialising the data with the function PSTestInit.

Usage

PSTestRun(
  proc_dat,
  formula,
  interaction = NULL,
  M = 1000,
  covariates = NULL,
  latent_effect,
  PS = "either",
  no_nn = 1,
  residual_tests = F,
  sigma = NULL,
  leaveoneout = T,
  fix_n = F,
  parallel = F,
  ncores = 1,
  simultaneous = F,
  global_alpha = 0.05,
  return_rho_vals = F,
  return_plots = T,
  return_model = F
)

Arguments

proc_dat	is the processed data from PSTestInit.
formula	is a formula in the R language describing the model to be fitted by spatstat. For spacetime data, this can be a named list of formulas. The name of each element of the list matches a name of one of the observed time steps. Each element gives the formula for the time step.
interaction	is an object of spatstat class "interact" describing the point process interaction structure, can also be a function that makes such an object, or NULL indicating that a Poisson process (stationary or nonstationary) should be fitted. See the help file for ppm from the spatstat package for more details.
M	specifies the number of Monte Carlo samples to take
covariates	is a list when discrete==F, whose entries are images, functions, windows, tessellations or single numbers when spatial. When spacetime, covariates is either a named list of lists (one per timestep), or a list of entries of covariates constant through time. When covariates change over time, name the lists with the number that points the list of covariates to the correct time. See the help file for ppm from the spatstat package for allowed types of covariates. When discrete==T, covariates is a data.frame object, in both the spatial and spacetime setting.
latent_effect	is the latent effect of interest. When discrete==F, it needs to be predicted over a high resolution grid. For discrete==T, the latent effect needs to be predicted at each areal unit. For discrete==T, this is a vector of class numeric. For discrete==T, the length must equal the number of areal units in the population multiplied by the number of unique time steps. For discrete==F, latent effect is of class im, or a SpatialPixels style format. For spacetime data when discrete==F, latent_effect is either a named list of im or SpatialPixels style objects (one per time period), or as before for spatial. The naming convention for the named list is the same as before.
PS	is a string specifying the direction of PS in the alternative. One of 'positive', 'negative' or 'either'.
no_nn	specifies the maximum number of nearest neighbours to average the distance over. Results of the tests of all values K=1:no_nn are returned.
residual_tests	is a logical argument specifying if the rank correlation of the smoothed raw HPP and model residuals should be computed.
sigma, leaveoneout	are additional arguments for the density function in spatstat.
fix_n	is a logical stating if the sample size should be fixed in the Monte Carlo samples to value observed in the data.
parallel	is a logical specifying if the code should be run in parallel.
ncores	specifies the number of cores to parallelize over if parallel=T.
simultaneous	is a logical specifying if the simultaneous test should be computed that corrects for multiple testing. This performs a rank envelope test.
global_alpha	is a number between 0 and 1 specifying the significance level of the simultaneous test.
return_rho_vals	is a logical stating if the raw rank correlations should be returned.
return_plots	is a logical stating if the ggplot object should be printed (if simultaneous = T).
return_model	is a logical stating if the fitted model object from spatstat, or mgcv should be returned.

Value

A named list containing a minimum of the empirical pointwise p values. Depending on the arguments specified above, simultaneous p values, rank correlations and plot objects may also be returned.

Examples

For detailed examples, see the vignette (i.e. run vignette('PSTestR')), or visit https://github.com/joenomiddlename/PStestR for more details.

See Also

PSTestInit

joenomiddlename/PStestR documentation built on March 26, 2022, 8:17 a.m.