ggdmc (version 0.1.3.9)

h.simulate.dmc: Simulate Choice-RT Data for Multiple Participants

Description

h.simulate.dmc generates random choice-RT responses based on an EAM model, which usually set up by model.dmc and a prior distribution settings, which usually created by prior.p.dmc.

Usage

h.simulate.dmc(object, nsim = 2, seed = NULL, ns = 1, ps = NA,
  SSD = Inf, p.prior = NA, staircase = NA, subject.cv = NULL)

Arguments

object

a DMC model

nsim

number of trials. Default is 2

seed

for comparible reason. Default is NULL.

ns

the number of subjects to be simulated. Default is 1

ps

a parameter x subject matrix

SSD

stop-signal parameter. is for use only with stop-signal designs, specified like n except a vector form is also allowed that is the same length as the data. It must have Inf in all go cells

p.prior

prior parameter list

staircase

staircase modifies SSD (see simulate.dmc).

subject.cv

a data frame, column names are in names(p.prior),

Details

The simulation engine of the diffusoin model calls rdiffusion in rtdists, which is adapted Voss & Voss' construct-sample C function in their fast-dm C software h.simulate.dmc behaves like simulate.dmc, but creates data for a set of ns subjects. It simulates based either on a fixed-effect model or on a random-effect model. For the latter, (i.e., hierarchical model), the user must supply a p.prior, in which case ps, (used like p.vector in simulate.dmc) is ignored. That is, h.simulate.dmc will generate a set of true EAM parameters stochastically based on the supplied prior distributions (stored in the user supplied p.prior). ps stands for "true" model parameters. It can be a vector exactly like p.vector, in which case each subject has identical parameters. All equal to the values in p.vector. ps can also be a matrix with one row per subject, in which case must have ns rows. It is saved (in expanded form) as "parameters" attribute in the generated data. p.prior is a list of distributions from which subject parameters are sampled. It is usually created by prior.p.dmc and will be saved as p.prior attribute. nsim can be a single number for a balanced design or matrix for an unbalanced design, where rows are subjects and columns are design cells. If the matrix has one row then all subjects have the same n in each cell, if it has one column then all cells have the same n, otherwise each entry specifies the n for a particular design subject x design cell combination.

Examples

## Set up a DDM Model, rate effect of factor F
m1 <- model.dmc(
 p.map     = list(a="1", v="F", z="1", d="1", sz="1", sv="1", t0="1",
 st0="1"),
 match.map = list(M=list(s1="r1", s2="r2")),
 factors   = list(S=c("s1","s2"), F=c("f1","f2")),
 constants = c(st0=0,d=0),
 responses = c("r1","r2"),
 type      = "rd")

## Population distribution
p.mean  <- c(a=2,   v.f1=2.5, v.f2=1.5, z=0.5, sz=0.3, sv=1,  t0=0.3)
p.scale <- c(a=0.5, v.f1=.5,  v.f2=.5,  z=0.1, sz=0.1, sv=.3, t0=0.05)
pop.prior <- prior.p.dmc(
  dists = rep("tnorm",7),
  p1    = p.mean,
  p2    = p.scale,
  lower = c(0,-5, -5, 0, 0, 0, 0),
  upper = c(5, 7,  7, 2, 2, 2, 2))

##  Check population distributions
plot_priors(pop.prior)

## Simulate some data
dat1 <- h.simulate.dmc(m1, nsim=20, ns=5, p.prior=pop.prior)
head(dat1)
##    S  F  R        RT
## 1 s1 f1 r1 0.9227881
## 2 s1 f1 r1 0.7878554
## 3 s1 f1 r1 0.4814711
## 4 s1 f1 r1 0.6864110
## 5 s1 f1 r1 0.5068179
## 6 s1 f1 r1 0.6356547

## Use LBA as an example
m2 <- model.dmc(
      p.map     = list(A="1", B="1", mean_v="M", sd_v="M", t0="1",
      st0="1"),
      match.map = list(M=list(s1=1, s2=2)),
      factors   = list(S=c("s1", "s2")),
      constants = c(st0= 0, sd_v.false=1),
      responses = c("r1", "r2"),
      type      = "norm")

## Population distribution
p.mean  <- c(A=.4,B=.6,mean_v.true=1,mean_v.false=0,sd_v.true = .5,t0=.3)
p.scale <- c(A=.1,B=.1,mean_v.true=.2,mean_v.false=.2,sd_v.true = .1,
t0=.05)
pop.prior <- prior.p.dmc(
    dists = c("tnorm","tnorm","tnorm","tnorm","tnorm","tnorm"),
    p1=p.mean, p2=p.scale,
    lower=c(0,0,NA,NA,0,.1),upper=c(NA,NA,NA,NA,NA,1))

## A data frame
dat2 <- h.simulate.dmc(m2, nsim=30, p.prior=pop.prior, ns=10) ##

## A 10-element list; each element is a participant
mdi2 <- data.model.dmc(dat2, m2)
head(mdi2[[1]])     ## Check the first participant (s1)
##    S  R        RT
## 1 s1 r1 0.7793124
## 2 s1 r2 0.5736594
## 3 s1 r2 0.6900489
## 4 s1 r2 2.3713993
## 5 s1 r1 1.5139890
## 6 s1 r2 0.5649499