witness.genetic: Fit the WITNESS model using a Genetic Algorithm

In dustinfife/witness: This package fits eyewitness data using Clark's (2003) WITNESS model

Description

After generating arbitrary starting values, the genetic algorithm selects the best parameter combinations to "mate." At each iteration, mutations are added with a specified probability, as well as totally new parameter values.

Usage

  witness.genetic(parameter.form, data = NULL, N = 1000,
    generations = 100, pcross = 0.8, mutProb = 0.001,
    prop.random = 0.001, gradient = 0.01, ...)

Arguments

parameter.form	a matrix where the rows are the experiments, and the columns are the parameters a, ssp, sfs, c, and wa. To fix parameters across rows, one would simply input the same values. For example, if one wanted all the encoding values (a) to be the same, the first row could consist of only ones.
data	a matrix that has the same number of rows as parameter.form. Again, the rows are the experiments but the columns correspond to 1. Target Identification, 2. Foil Identification, and 3. No Identification.
N	the number of iterations the WITNESS model will do to simulate the eyewitness procedure.
generations	how many generations the genetic algorithm will iterate through before quitting.
pcross	probability of cross "breeding," or the probability that a parent will breed.
mutProb	probability of randomly mutating a parameter (i.e., the probability that noise will be added to the next generation)
prop.random	to avoid local minima, the algorithm allows for "aliens" to be introduced (or completely new parameter values).
gradient	convergence criteria. Currently not implemented.
...	other parameters passed to the WITNESS model.

References

Clark, S. E. (2003). A memory and decision model for eyewitness identification. Applied Cognitive Psychology, 17, 629-654.

See Also

witness.optim, witness.est

Examples

dataMatrix = matrix(c(.471, .230, .350,
							.208, .137, .513,
							.396, .431, .242,
							.166, .081, .669), nrow=4, byrow=TRUE)
# create an parameter.form
params = matrix(c(rep("e", times=4),
		0, 0, 0, 0,
		"sfs1", "sfs1", "sfs2", "sfs2",
		"cr1", "cr2", "cr1", "cr2",
		1, 1, 1, 1), nrow=4)

		####### do genetic algorithm (commented b/c it takes a while)
#fit = witness.genetic(params, data=dataMatrix, N=10, generations=1, sample.size=100, meth="WITC")

dustinfife/witness documentation built on May 15, 2019, 6:05 p.m.