pbcm.du: Data-uninformed Parametric Bootstrap Cross-fitting
In hkauhanen/pbcm: Parametric Bootstrap Cross-Fitting Method

Description Usage Arguments Details Value Author(s) References Examples

View source: R/pbcm.du.R

The data-uninformed Parametric Bootstrap Cross-fitting Method (PBCM) generates synthetic data from two models of a phenomenon with given model parameter values, and then cross-fits the models to these data. The result is two distributions of the goodness of fit metric Δ GoF = GoF_1 - GoF_2, where GoF_1 is the fit of model 1 and GoF_2 the fit of model 2.

pbcm.du(
  fun1,
  fun2,
  genfun1,
  genfun2,
  reps,
  args1 = NULL,
  args2 = NULL,
  genargs1 = NULL,
  genargs2 = NULL,
  print_genargs = TRUE,
  verbose = TRUE,
  GoFname = "GoF"
)

`fun1`	First modelling function
`fun2`	Second modelling function
`genfun1`	Generator function for first model
`genfun2`	Generator function for second model
`reps`	Number of Monte Carlo repetitions
`args1`	List of arguments passed to `fun1`
`args2`	List of arguments passed to `fun2`
`genargs1`	List of arguments passed to `genfun1`
`genargs2`	List of arguments passed to `genfun2`
`print_genargs`	Whether the generator argument values should be included in output (see Details)
`verbose`	If `TRUE`, a progress bar is printed to the console and warnings are issued
`GoFname`	Name of the element returned by `fun1` and `fun2` holding the goodness of fit; see Details

Functions fun1 and fun2 must take an argument named data in addition to any arguments specified in args1 and args2; this is used to pass the synthetic data generated by genfun1 and genfun2. Moreover, these functions must return a list with at least one element carrying the goodness of fit; the name of this element may be specified through the GoFname argument, by default the string "GoF" is assumed.

A data frame in long format with the following columns:

rep: Monte Carlo repetition number
generator: Generating model
GoF1: Goodness of fit of model 1
GoF2: Goodness of fit of model 2
DeltaGoF: Equals GoF1 - GoF2

In addition to these columns, if print_genargs == TRUE, each argument in the lists genargs1 and genargs2 is included as a column of its own, with the argument's name prefixed by "genargs1_" or "genargs2_".

Henri Kauhanen

Wagenmakers, E.-J., Ratcliff, R., Gomez, P. & Iverson, G. J. (2004) Assessing model mimicry using the parametric bootstrap. Journal of Mathematical Psychology, 48(1), 28–50. https://doi.org/10.1016/j.jmp.2003.11.004

x <- seq(from=0, to=1, length.out=100)
mockdata <- data.frame(x=x, y=x + rnorm(100, 0, 0.5))

myfitfun <- function(data, p) {
  res <- nls(y~a*x^p, data, start=list(a=1.1))
  list(a=coef(res), GoF=deviance(res))
}

mygenfun <- function(a, p) { 
  x <- seq(from=0, to=1, length.out=100)
  y <- a*x^p + rnorm(100, 0, 0.5)
  data.frame(x=x, y=y)
}

pbcm.du(fun1=myfitfun, fun2=myfitfun, genfun1=mygenfun, genfun2=mygenfun, 
        reps=20, args1=list(p=1), args2=list(p=2), 
        genargs1=list(a=1.1, p=1), genargs2=list(a=1.1, p=2))

sweep <- lapply(X=seq(from=0.5, to=1.5, by=0.1),
                FUN=function(X) {
                  pbcm.du(fun1=myfitfun, fun2=myfitfun, genfun1=mygenfun,
                          genfun2=mygenfun, reps=20,
                          args1=list(p=1), args2=list(p=2), 
                          genargs1=list(a=X, p=1), genargs2=list(a=X, p=2))
                })

sweep <- do.call(rbind, sweep)

sweep$parameter <- ifelse(is.na(sweep$genargs1_a), sweep$genargs2_a, sweep$genargs1_a)

## Not run: 
  library(ggplot2)
  g <- ggplot(sweep, aes(x=DeltaGoF, fill=generator)) + geom_density(alpha=0.5) 
  g <- g + facet_wrap(.~parameter)
  print(g)

## End(Not run)