ddmc: Compute Probability Density of Drift-Diffusion Model

In TasCL/ggdmc: Dynamic Model of Choice with Parallel Computation, and C++ Capabilities

Description

This function implements the equations in Voss, Rothermund, and Voss (2004). These equations calculate Ratcliff's drift-diffusion model (RDM, 1978). ddmc re-implements Voss, Rothermund, and Voss's (2004) equations A1 to A4 (page 1217) via Rcpp. This Rcpp function is akin to DMC's likelihood.dmc. There is a ddmc_parallel with same arguments using Open MPI to calculate RDM probability densities when data sets are large (e.g., more than 5000 trials per condition) or when precision is set high.

Usage

ddmc(x, pVec, precision = 2.5, minLike = 1e-10)

ddmc_parallel(x, pVec, precision = 2.5, minLike = 1e-10)

Arguments

x	a data frame containing choices and RTs. This is akin to dnorm's x argument.
pVec	a parameter vector. For example, p.vector <- c(a=1.25, v.f1=.20, v.f2=.33, z=.67, sz=.26, sv=.67, t0=.26)
precision	a precision parameter. The larger, the higher precision to estimate diffusion probability density. A general recommendation for the low bound of precision is 2.5. If you need to use a precision higher than that, you may want to consider using ddmc_parallel. A default value is set at 2.5.
minLike	a minimal log likelihood. If a estimated density is less than minLike, the function returns minLike. A default value is set at 1e-10.

Value

a double vector with drift-diffusion probability density

References

Voss, A., Rothermund, K., & Voss, J. (2004). Interpreting the parameters of the diffusion model: An empirical validation. Memory & Cognition, 32(7), 1206-1220.

Ratcliff, R. (1978). A theory of memory retrival. Psychological Review, 85, 238-255.

Examples

## Set up a basic RDM design
m1 <- model.dmc(
    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")

pVec <- c(a=1, v=1, z=.5, sz=.25, sv=.2,t0=.15)

## Set up a model-data instance
raw.data <- simulate(m1, nsim=1e2, p.vector=pVec)
mdi <- data.model.dmc(raw.data, m1)

## Print probability densities on the screen
ddmc(mdi, pVec)
## [1]  0.674039828 0.194299671 0.377715949 1.056043402 0.427508806
## [6]  2.876521367 0.289301835 2.701315781 0.419909813 1.890380685
## [11] 2.736803321 1.278188278 0.607926999 0.024858130 1.878312335
## [16] 0.826923114 0.189952016 1.758620294 1.696875882 2.191290074
## [21] 2.666750382 ...

mdi.large <- data.model.dmc(simulate(m1, nsim=1e4, p.vector=pVec), m1)

## d.large <- dplyr::tbl_df(mdi.large)
## d.large
## Source: local data frame [20,000 x 3]
##
##         S      R        RT
##    (fctr) (fctr)     (dbl)
## 1      s1     r2 0.2961630
## 2      s1     r1 0.9745674
## 3      s1     r2 0.6004945
## 4      s1     r1 0.1937356
## 5      s1     r1 0.3608844
## 6      s1     r1 0.5343379
## 7      s1     r1 0.2595707
## 8      s1     r1 0.3247941
## 9      s1     r1 0.4901325
## 10     s1     r2 0.5011935
## ..    ...    ...       ...

## system.time(den1 <- ddmc(mdi.large, pVec))
##  user  system elapsed
## 0.028   0.004   0.046

## system.time(den2 <- ddmc_parallel(mdi.large, pVec))
##  user  system elapsed
## 0.200   0.000   0.023
## all.equal(den1, den2)
## [1] TRUE

TasCL/ggdmc documentation built on May 9, 2019, 4:19 p.m.