ggdmc (version 0.2.5.2)

profile.model: Profile a model object

Description

This function produces data for profiling model likelihoods based on a data model instance (ie fitted).

Usage

# S3 method for model
profile(fitted, pname, minp, maxp, p.vector,
  npoint = 100, digits = 2, ylim = NA, nthread = 32, ...)

Arguments

fitted

a data model instance

pname

indicate which parameter in theta to plot. For example, in a LBA model with a pVec <- c(A="1",B="1",mean_v="M",sd_v="1",t0="1", st0="1"), one can assign pname <- "A" to ask profile to profile A parameter

minp

lower bound for pname parameter

maxp

upper bound for pname parameter

p.vector

a parameter vector. Use Lognromal LBA model as an example, pVec <- c(A = .25, B = .35, meanlog_v.true = 1, meanlog_v.false = .25, sdlog_v.true = .5, t0 = .2)

npoint

grid for p.name parameter

digits

print out how many digits

ylim

y range

nthread

number of thread in a GPU block. This argument works for GPU-based PDA probability density functions)

...

additional optional arguments.

Details

The argument, pname indicates which model parameter to profile. For example, if we want to profile the boundary separation a in a DDM, we extract it from a p.vector and calculates its marginal likelihoods based on the data model instance on a grid of n.point.

Briefly, the function sets a range for the points on the x axis, which is the values for the profiled parameter (e.g., a). Then it initiates a log-likelihodd vector with length of n.point and 0 everywhere. Next, it keeps the other parameters in the model fixed and changes only the target parameter values to ps[i]. After that, it calculates their sum log-likelihoods. Finally, it stores the log-likelihoods in the i position of the ll vector.

Examples

# NOT RUN {
model <- BuildModel(
        p.map     = list(a = "1",v = "1",z = "1", d = "1", sz = "1",
                    sv = "1", t0 = "1", st0 = "1"),
        constants = c(st0 = 0, d = 0),
        match.map = list(M = list(s1 = "r1", s2 = "r2")),
        factors   = list(S = c("s1", "s2")),
        responses = c("r1", "r2"),
        type      = "rd")

p.prior <- BuildPrior(
  dists = rep("tnorm", 6),
  p1=c(a=2,   v=2.5, z=0.5, sz=0.3, sv=1,  t0=0.3),
  p2=c(a=0.5, v=.5,  z=0.1, sz=0.1, sv=.3, t0=0.05),
  lower=c(0,-5, 0, 0, 0, 0),
  upper=c(5, 7, 2, 2, 2, 2))

p.vector <- c(a=1,v=1, z=0.5, sz=0.25, sv=0.2,t0=.15)
dat <- simulate(model, 1e2, ps = p.vector)
dmi <- BuildDMI(dat, model)

## ------------------------------40
par(mfrow=c(2,3));
profile(dmi, "a",  .1,   2, p.vector)
profile(dmi, "v",  .1,   2, p.vector)
profile(dmi, "z",  .2,  .8, p.vector)
profile(dmi, "sz", .1,  .9, p.vector)
profile(dmi, "sv", .1,   2, p.vector)
profile(dmi, "t0", .01, .5, p.vector)
par(mfrow=c(1,1));
# }