Description Usage Arguments Details Value References Examples
Constructs a Generalized Context Model.
1 2 |
learning |
an object of class |
response |
an object of class |
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. |
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 similarity between two cues is computed as
s(x[t],x[k]) = exp(-lambda * d(x[t],x[k])^(q) )
where d(x[t],x[k]) is the generalized Minkowski distance
d(x[t],x[k]) = (sum[j] w[j] | x[jt] - x[jk] |^(r))^(1/r)
When making a response to cue x[t], first the overall similarity to each of the previously encountered exemplars of each of the levels of outcome y is determined as
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.
A more general version of the GCM is implemented by gGCM
.
An object of class GCM
.
Nosofsky, R.M. (1986). Attention, Similarity, and the Identification-Categorization Relationship. Journal of Experimental Psychology: General, 115, 39-57.
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 <- GCM(y~x1+x2+x3+x4,response=r~1,data=controls,
fix=list(r=TRUE,q=TRUE),ntimes=rep(200,16),
remove.intercept=TRUE)
## Not run:
## estimate free parameters
## discount (ignore) first 5 responses in each series
## as these can give deterministic predictions
mod <- fit(mod,discount=5)
## End(Not run)
summary(mod)
|
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.