Description Usage Arguments Details Value References Examples
Constructs a generalized version of the Generalized Context Model.
1 2 3 4 5 | gGCM(learning,response,level=c("nominal","interval"),distance=c("cityblock",
"euclidian","minkowski"),similarity=c("exponential","gaussian","general"),
sampling=c("uniform","power","exponential"),parameters=list(w=NULL,lambda=1,
r=1,q=1,gamma=NULL,theta=NULL,sdy=NULL,sdr=NULL),fixed,data,subset,ntimes=NULL,
replicate=TRUE,base=NULL,remove.intercept=FALSE)
|
learning |
an object of class |
response |
an object of class |
level |
the measurement level of the dependent variable, either nominal or interval. |
distance |
either the name of a standard distance function, or a function which returns a T*T matrix with distances between the cues. See details. |
similarity |
either the name of a standard similarity function, or a function which converts the T*T distance matrix to a T*T similarity matrix. See datails. |
sampling |
either the name of a standard sampling function, or a function which returns a T*T matrix with sampling weights. See details. |
parameters |
an (optional) named list with (starting) values of the parameters. If no values are supplied, defaults are used. |
fixed |
a logical vector indicating whether model parameters are fixed or free |
data |
data frame containing the variables listed in the formula argument. |
subset |
subset of data to fit model to. |
ntimes |
a vector with number of observations for each replication. |
replicate |
logical to indicate whether model parameters are identical for each replication in ntimes. |
base |
assign one of the levels of the criterion variable the role of base category. |
remove.intercept |
(logical) should the intercept term be removed from the x matrix of the model? |
The Generalized Context Model (Nosofsky, 1986) is an exemplar model. It predicts the value of a criterion at t based on the similarity of a probe cue x[t] to stored cues x[t-k], k = 1,...,t-1.
The GCM can be seen as a mixture model. Each encountered exemplar adds a new component to the mixture. The mixture proportions are defined by the similarities. See the package manual for more information.
The model implemented by gGCM
extends the original GCM (Nosofsky, 1986)
by allowing (1) a continuous criterion, and (2) memory decay of exemplars.
An object of class gGCMinterval
or gGCMnominal
.
Nosofsky, R.M. (1986). Attention, Similarity, and the Identification-Categorization Relationship. Journal of Experimental Psychology: General, 115, 39-57.
Speekenbrink, M. \& Shanks, D.R. (2010). Learning in a changing environment. Journal of Experimental Psychology: General.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 | ## open weather prediction data
data(WPT)
controls <- subset(WPT,id %in% paste("C",1:16,sep=""))
## initialize model, use remove.intercept=TRUE so that the x matrix will contain
## four columns.
mod <- gGCM(y~x1+x2+x3+x4,response=r~1,data=controls,ntimes=rep(200,16),remove.intercept=TRUE)
## estimate free parameters
## Not run: mod <- fit(mod)
summary(mod)
# now a model with a general minkowski distance function, and an exponential
# memory decay. Using numerical predictors and removing the intercept (-1) in
# the model formula removes the need for the remove.intercept argument.
mod <- gGCM(y~as.numeric(x1==1)+as.numeric(x2==1)+as.numeric(x3==1)+
as.numeric(x4==1)-1,response=r~1,distance="minkowski",sampling="uniform",data=controls,
ntimes=rep(200,16))
|
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.