simulate.scampr: Simulate functiion for objects of class scampr

In ElliotDovers/scampr: Spatially Correlated, Approximate Modelling of Point patterns in R

Simulate functiion for objects of class scampr

Description

Simulates point patterns from a specified scampr model.

Usage

## S3 method for class 'scampr'
simulate(
  object,
  nsim = 1,
  seed = NULL,
  ...,
  domain.data,
  rcoef.density = c("posterior", "prior"),
  which.intensity = c("expected", "sample"),
  return.type = c("data.frame", "ppp"),
  nsurv,
  log.expected = T
)

Arguments

object	a scampr model
nsim	number of point patterns to simulate. Defaults to 1.
seed	an integer for setting the random seed.
...	NA
domain.data	optionally, a data frame of point locations that adequately cover the domain of interest, as well as, predictors involved in the model. If missing model quadrature is used. NOTE: if model quadrature is used and is an inadequate cover of the domain, simulation may be effected as interpolated variables will contain many NA values.
rcoef.density	a character string indicating the probability density of the random effects to draw from.
which.intensity	a character string, one of 'expected' or 'sample', indicating whether to simulate from the expected intensity (with respect to 'type' and 'rcoef.density') or from a randomly sampled intensity (with respect to 'rcoef.density') respectively. This will work in conjunction with 'type' (changes expectation) and 'rcoef.density' (N(0,prior_var) or N(posterior_mean, posterior_var)). Toggling this depends on the use of simulation, e.g. 'expected' may be useful in validating a model while 'sample' may be useful for making predictions.
return.type	a character string, one of 'data.frame' or 'ppp', indicating the object type of the simulation to be returned. Default is data frame for use in scampr models. ppp object is useful for interfacing with spatstat.core::envelope e.g.
nsurv	an optional integer describing the number of survey sites to be included in the simulated presence/absence data. Only applies to combined data models (popa), if null the survey sites of the original model are used.
log.expected	a logical indicating whether to take the expectation of the log-intensity (ignores the variance correction). Only relevant to LGCP models for which which.intensity is 'expected'.

Value

Depends on return.type. Default is a point pattern object of class 'ppp' from spatstat. Otherwise can be set to return a data.frame describing the point pattern (with corresponding quadrature as an attribute). If nsim > 1 then returned as a list of either 'return.type'.

Examples

dat <- scampr::gorillas
dat$elev <- scale(dat$elevation)
mod <- scampr(pres ~ elev, dat, include.sre = F)
## Not run: pp <- simulate(mod)

ElliotDovers/scampr documentation built on March 17, 2024, 3:27 p.m.