ggdmc (version 0.1.3.9)

h.run.dmc: Fit an EAM with Multiple Participants

Description

h.run.dmc calls run_data in the case of fixed-effect model, or run_hyper in the case of random-effect model. At the moment, it fits either fixed-effect or random-effect model by looking for the hyper attribute in a DMC sample/object.

Usage

h.run.dmc(samples, report = 100, cores = 1, blocks = NA, p.migrate = 0,
  h.p.migrate = 0, force.hyper = NA, force.data = NA, gamma.mult = 2.38,
  h.gamma.mult = NA, setting = NULL, verbose = FALSE)

Arguments

samples

a DMC samples, usually generated by calling h.samples.dmc.

report

the progress report. Default is every 100 iterations report once

cores

a switch for parallel computation. In the case of fixed-effect model, h.run.dmc calls snow (via snowfall) or parallel. When setting for example cores=4, h.run.dmc requests 4 parallel R processes. In the case of random-effect model, h.run.dmc calls OpenMP library, which will automatically decide the number of cores to use, so setting core number higher than 1 has the same effect. Use OpenMP only when the data set is large (e.g., more than 1000 trial per condition).

blocks

a DEMCMC parameter for blocking. Not implemented yet. Default is NA

p.migrate

the probability of applying migration sampler. Set it greater than will trigger the migration sampler. For example p.migrate=0.05 will use migration sampler about 5% chance.

h.p.migrate

similar to p.migrate for hyper level

force.hyper

This argument has no funciton. It is here only for compatibility (with DMC) reason.

force.data

This argument has no funciton. It is here only for compatibility (with DMC) reason.

gamma.mult

a DE-MCMC tuning parameter, affecting the size of jump. Default is 2.38.

h.gamma.mult

Similar with gamm.mult. This argument has no funciton. It is here only for compatibility reason.

setting

a list to store setting arguments, such as p.migrate, gamma.mult, etc, sending to C++ function. The user should not use this argument

verbose

a swtich to see debugging information

Examples

m1 <- model.dmc(
p.map     = list(a="1",v="1",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")),
constants = c(st0=0, d=0),
responses = c("r1","r2"),
type      = "rd")

## Population distribution
pop.prior <- prior.p.dmc(
  dists = rep("tnorm", 6),
  p1    = c(a=2,  v=2.5, z=.5, sz=.3, sv=1,  t0=.3),
  p2    = c(a=.5, v=.5,  z=.1, sz=.1, sv=.3, t0=.05),
  lower = c(0,-5, 0, 0, 0, 0),
  upper = c(5, 7, 2, 2, 2, 2))

dat <- h.simulate.dmc(m1, nsim=30, p.prior=pop.prior, ns=8)
mdi <- data.model.dmc(dat, m1)
p.prior  <- prior.p.dmc(
  dists = rep("tnorm", 6),
  p1    = c(a=2,  v=2.5, z=.5, sz=.3, sv=1,  t0=.3),
  p2    = c(a=.5, v=.5,  z=.1, sz=.1, sv=.3, t0=.05) * 5,
  lower = c(0,-5, 0, 0, 0, 0),
  upper = c(5, 7, 2, 2, 2, 2))

## Fixed-effect model
samplesInit <- h.samples.dmc(nmc=20, p.prior=p.prior, data=mdi, thin=1)
samples0    <- h.run.dmc(samples=samplesInit, report=10)

## Make a hyper-prior list
mu.prior <- prior.p.dmc(
  dists = rep("tnorm", 6),
  p1    = c(a=2,  v=2.5, z=.5, sz=.3, sv=1,  t0=.3),
  p2    = c(a=.5, v=.5,  z=.1, sz=.1, sv=.3, t0=.05) * 5,
  lower = c(0,-5, 0, 0, 0, 0),
  upper = c(5, 7, 2, 2, 2, 2))

sigma.prior <- prior.p.dmc(
  dists = rep("beta", 6),
  p1    = c(a=1, v=1, z=1, sz=1, sv=1, t0=1),
  p2    = c(1,1,1,1,1,1),
  upper = c(2,2,2,2,2,2))

pp.prior <- list(mu.prior, sigma.prior)

## Random-effect model
hsamplesInit <- h.samples.dmc(nmc=20, p.prior=p.prior, pp.prior=pp.prior,
  data=mdi, thin=1)
hsamples0 <- h.run.dmc(samples=hsamplesInit, report=10, p.migrate=.05,
  h.p.migrate=.05)