A framework for the indirect classification approach.
Usage
## S3 method for class 'data.frame'
inclass(formula, data, pFUN = NULL, cFUN = NULL, ...)
Arguments
formula
formula. A formula specified as
y~w1+w2+w3~x1+x2+x3 models each intermediate variable
w1, w2, w3 by wi~x1+x2+x3 and the response by
y~w1+w2+w3 if no other formulas are given in pFUN or cFUN.
data
data frame of explanatory, intermediate and response variables.
pFUN
list of lists, which describe models for the intermediate variables, see below for details.
cFUN
either a function or a list which describes the model for the
response variable. The function has the argument newdata only.
...
additional arguments, passed to model fitting of the
response variable.
Details
A given data set is subdivided into three types of variables: those to be
used predicting the class (explanatory variables) those to be used defining
the class (intermediate variables) and the class membership variable itself
(response variable). Intermediate variables are modelled based on the
explanatory variables, the class membership variable is defined on the
intermediate variables.
Each specified intermediate variable is modelled separately
following pFUN and a formula specified by formula.
pFUN is a list of lists, the maximum length of
pFUN is the number of intermediate variables. Each element of
pFUN is a list with elements: model - a function with arguments formula and
data; predict - an optional function with arguments object, newdata only,
if predict is not specified, the predict method of model
is used; formula - specifies the formula for the corresponding
model (optional),
the formula described in y~w1+w2+w3~x1+x2+x3 is used if no other is
specified.
The response is classified following cFUN, which is either a fixed
function or a list as described below. The determined function cFUN assigns the intermediate (and
explanatory) variables to a certain class membership, the list
cFUN has the elements formula, model, predict and
training.set. The elements formula, model, predict are
structured as described by pFUN, the described model is
trained on the original (intermediate variables) if training.set="original"
or if training.set = NULL, on the fitted values if
training.set = "fitted" or on observations not included in a
specified subset if training.set = "subset".
A list of prediction models corresponding to each
intermediate variable, a predictive function for the response, a list of
specifications for the intermediate and for the response are returned.
For a detailed description on indirect
classification see Hand et al. (2001).
Value
An object of class inclass, consisting of a list of
model.intermediate
list of fitted models for each intermediate
variable.
model.response
predictive model for the response variable.
para.intermediate
list, where each element is again a list and specifies
the model for each intermediate variable.
para.response
a list which specifies the model for response variable.
References
David J. Hand, Hua Gui Li, Niall M. Adams (2001),
Supervised classification with structured class definitions.
Computational Statistics & Data Analysis36,
209–225.
Andrea Peters, Berthold Lausen, Georg Michelson and Olaf Gefeller (2003),
Diagnosis of glaucoma by indirect classifiers.
Methods of Information in Medicine1, 99-103.
See Also
bagging, inclass
Examples
data("Smoking", package = "ipred")
# Set three groups of variables:
# 1) explanatory variables are: TarY, NicY, COY, Sex, Age
# 2) intermediate variables are: TVPS, BPNL, COHB
# 3) response (resp) is defined by:
classify <- function(data){
data <- data[,c("TVPS", "BPNL", "COHB")]
res <- t(t(data) > c(4438, 232.5, 58))
res <- as.factor(ifelse(apply(res, 1, sum) > 2, 1, 0))
res
}
response <- classify(Smoking[ ,c("TVPS", "BPNL", "COHB")])
smoking <- data.frame(Smoking, response)
formula <- response~TVPS+BPNL+COHB~TarY+NicY+COY+Sex+Age
inclass(formula, data = smoking, pFUN = list(list(model = lm, predict =
mypredict.lm)), cFUN = classify)