ddModel (version 0.2.9.0)

simulate,ddm-method: Simulate Data from a Decision Diffusion Model

Description

Simulate response times and choices from a Decision Diffusion model (DDM) associated with a specific experimental design. The specification is typically specified, using ggdmcModel::BuildModel).

Usage

# S4 method for ddm
simulate(
  object,
  nsim = 4L,
  seed = NULL,
  n_subject = 3L,
  parameter_vector = NULL,
  time_parameters = c(t_min = -1, t_max = 1, dt = 0.001),
  debug = FALSE
)

Value

A data frame with simulated trial data, including columns like subject, trial, choice, and rt.

Arguments

object

An object of class "ddm" that defines the model structure and parameters.

nsim

Integer. Number of trials to simulate per subject. Defaults to 4. The function adjusts nsim if not divisible by number of conditions.

seed

Optional integer. Sets the random seed for reproducibility. Defaults to NULL.

n_subject

Integer. Number of subjects to simulate. Defaults to 3.

parameter_vector

Optional. A named vector or list of parameters (e.g., a, z, v, t0) used as the true value to simulate. The function by default generates one or many true parameter vector based on the population distribution stored in the ddm object.

time_parameters

Numeric vector of for setting the time range and time step for simulation (using inverse transform method).

debug

Logical. If TRUE, print debugging output during simulation. Defaults to FALSE.

Details

This method simulates data from a design-based Decision Diffusion model (DDM). You can simulate multiple subjects, override default parameters, and use conduct inverse sampling with a fine time step. You may turn on debugging output when something goes wrong.

Examples

if (requireNamespace("ggdmcModel", quietly = TRUE)) {
  BuildModel <- getFromNamespace("BuildModel", "ggdmcModel")
  model <- ggdmcModel::BuildModel(
    p_map = list(
      a = "1", v = "1", z = "1", d = "1", sz = "1", sv = "1",
      t0 = "1", st0 = "1", s = "1", precision = "1"
    ),
    match_map = list(M = list(s1 = "r1", s2 = "r2")),
    factors = list(S = c("s1", "s2")),
    constants = c(d = 0, s = 1, st0 = 0, sv = 0, precision = 3),
    accumulators = c("r1", "r2"),
    type = "fastdm"
  )

# Simulate one subject ------------------
sub_model <- setDDM(model)
p_vector <- c(a = 1, sz = 0.25, t0 = 0.15, v = 2.5, z = .38)
dat <- simulate(sub_model,
  nsim = 256, parameter_vector = p_vector,
  n_subject = 1
)
# Simulate multiple subjects ------------------
# 1. First build a population distribution to sample many p_vector's

pop_mean <- c(a = 1, sz = 0.25, t0 = 0.15, v = 2.5, z = 0.38)
pop_scale <- c(a = 0.05, sz = 0.01, t0 = 0.02, v = .5, z = 0.01)
pop_dist <- ggdmcPrior::BuildPrior(
  p0    = pop_mean,
  p1    = pop_scale,
  lower = c(0, 0, 0, -10, 0),
  upper = rep(NA, model@npar),
  dists = rep("tnorm", model@npar),
  log_p = rep(FALSE, model@npar)
)

# 2. Enter the population distribution to a DDM model
pop_model <- setDDM(model, population_distribution = pop_dist)

# 3. simulate method will use the distribution to sample new p_vector's
# Note: You must enter sensible pop_mean, pop_scale and distribution.
hdat <- simulate(pop_model, nsim = 128, n_subject = 32)
}