Title:	Tools and Classes for Statistical Models
Date:	2025-05-02
Version:	0.2-24
Description:	A collection of tools to deal with statistical models. The functionality is experimental and the user interface is likely to change in the future. The documentation is rather terse, but packages ‘coin’ and ‘party’ have some working examples. However, if you find the implemented ideas interesting we would be very interested in a discussion of this proposal. Contributions are more than welcome!
Depends:	stats, stats4
Imports:	methods
LazyLoad:	yes
License:	GPL-2
NeedsCompilation:	no
Packaged:	2025-05-02 09:58:16 UTC; hothorn
Author:	Torsten Hothorn [aut, cre], Friedrich Leisch [aut], Achim Zeileis [aut]
Maintainer:	Torsten Hothorn <Torsten.Hothorn@R-project.org>
Repository:	CRAN
Date/Publication:	2025-05-02 14:30:07 UTC

Class "FormulaParts"

Description

A class describing the parts of a formula.

Objects from the Class

Objects can be created by calls of the form new("FormulaParts", ...).

Slots

formula:

Object of class "list".

Methods

No methods defined with class "FormulaParts" in the signature.

Generic Utility Functions

Description

A collection of standard generic functions for which other packages provide methods.

Usage

ICL(object, ...)
KLdiv(object, ...)
Lapply(object, FUN, ...)
clusters(object, newdata, ...)
getModel(object, ...)
parameters(object, ...)
posterior(object, newdata, ...)
prior(object, ...)
refit(object, newdata, ...)
relabel(object, by, ...)
ParseFormula(formula, data = list())

Arguments

Details

ICL:

Integrated Completed Likelihood criterion for model selection.

KLdiv:

Kullback-Leibler divergence.

Lapply:

S4 generic for lapply

clusters:

Get cluster membership information from a model or compute it for new data.

getModel:

Get single model from a collection of models.

parameters:

Get parameters of a model (similar to but more general than coefficients).

posterior:

Get posterior probabilities from a model or compute posteriors for new data.

prior:

Get prior probabilities from a model.

refit:

Refit a model (usually to obtain additional information that was not computed or stored during the initial fitting process).

relabel:

Relabel a model (usually to obtain a new permutation of labels in mixture models or cluster objects).

Author(s)

Friedrich Leisch

Apply functions to Data in Object of Class "ModelEnv"

Description

Apply a single function or a collection of functions to the data objects stored in a model environment.

Usage

## S4 method for signature 'ModelEnv'
MEapply(object, FUN, clone = TRUE, ...)

Arguments

Examples

data("iris")
me <- ModelEnvFormula(Species+Petal.Width~.-1, data=iris,
                      subset=sample(1:150, 10))

me1 <- MEapply(me, FUN=list(designMatrix=scale,
                            response=function(x) sapply(x, as.numeric)))

me@get("designMatrix")
me1@get("designMatrix")

Class "ModelEnv"

Description

A class for model environments.

Details

Objects of class ModelEnv basically consist of an environment for data storage as well as get and set methods.

na.fail returns FALSE when at least one missing value occurs in object@env. na.pass returns object unchanged and na.omit returns a copy of object with all missing values removed.

Objects from the Class

Objects can be created by calls of the form new("ModelEnv", ...).

Slots

env:

Object of class "environment".

get:

Object of class "function" for extracting objects from environment env.

set:

Object of class "function" for setting object in environment env.

hooks:

A list of hook collections.

Methods

clone

signature(object = "ModelEnv"): copy an object.

dimension

signature(object = "ModelEnv", which = "character"): get the dimension of an object.

empty

signature(object = "ModelEnv"): Return TRUE, if the model environment contains no data.

has

signature(object = "ModelEnv", which = "character"): check if an object which is available in env.

initialize

signature(.Object = "ModelEnv"): setup new objects.

show

signature(object = "ModelEnv"): show object.

subset

signature(x = "ModelEnv"): extract subsets from an object.

na.pass

na.action method for ModelEnv objects.

na.fail

na.action method for ModelEnv objects.

na.omit

na.action method for ModelEnv objects.

Examples

### a new object
me <- new("ModelEnv")

## the new model environment is empty
empty(me)

### define a bivariate response variable
me@set("response", data.frame(y = rnorm(10), x = runif(10)))
me

## now it is no longer empty
empty(me)

### check if a response is available
has(me, "response")

### the dimensions
dimension(me, "response")

### extract the data
me@get("response")

df <- data.frame(x = rnorm(10), y = rnorm(10))

## hook for set method:
mf <- ModelEnvFormula(y ~ x-1, data = df, setHook=list(designMatrix=scale))
mf@get("designMatrix")
mf@set(data=df[1:5,])
mf@get("designMatrix")

### NA handling
df$x[1] <- NA
mf <- ModelEnvFormula(y ~ x, data = df, na.action = na.pass)
mf
na.omit(mf)

Generate a model environment from a classical formula based interface.

Description

A flexible implementation of the classical formula based interface.

Usage

ModelEnvFormula(formula, data = list(), subset = NULL, 
                na.action = NULL, frame = NULL, 
                enclos = sys.frame(sys.nframe()), other = list(), 
                designMatrix = TRUE, responseMatrix = TRUE,
                setHook = NULL, ...)

Arguments

Details

This function is an attempt to provide a flexible infrastucture for the implementation of classical formula based interfaces. The arguments formula, data, subset and na.action are well known and are defined in the same way as in lm, for example.

ModelEnvFormula returns an object of class ModelEnvFormula-class - a high level object for storing data improving upon the capabilities of data.frames.

Value

An object of class ModelEnvFormula-class.

Examples

### the `usual' interface
data(iris)
mf <- ModelEnvFormula(Species ~ ., data = iris)
mf

### extract data from the ModelEnv object
summary(mf@get("response"))
summary(mf@get("input"))
dim(mf@get("designMatrix"))

### contrasts
mf <- ModelEnvFormula(Petal.Width ~ Species, data = iris, 
                      contrasts.arg = list(Species = contr.treatment))
attr(mf@get("designMatrix"), "contrasts")
mf <- ModelEnvFormula(Petal.Width ~ Species, data = iris, 
                      contrasts.arg = list(Species = contr.sum))
attr(mf@get("designMatrix"), "contrasts")

### additional formulae
mf <- ModelEnvFormula(Petal.Width ~ Species, data = iris, 
                      other = list(pl = ~ Petal.Length))
ls(mf@env)
identical(mf@get("pl")[[1]], iris[["Petal.Length"]])

Class "ModelEnvFormula"

Description

A class for formula-based model environments.

Objects from the Class

Objects can be created by calls of the form new("ModelEnvFormula", ...).

Slots

env:

Object of class "environment".

get:

Object of class "function" for extracting objects from environment env.

set:

Object of class "function" for setting object in environment env.

formula:

Object of class "list".

hooks:

A list of hook collections.

Extends

Class "ModelEnv", directly. Class "FormulaParts", directly.

Methods

No methods defined with class "ModelEnvFormula" in the signature.

Generate a model environment from design and response matrix

Description

A simple model environment creator function working off matrices for input and response. This is much simpler and more limited than formula-based environments, but faster and easier to use, if only matrices are allowed as input.

Usage

ModelEnvMatrix(designMatrix=NULL, responseMatrix=NULL,
               subset = NULL, na.action = NULL, other=list(), ...)

Arguments

Details

ModelEnvMatrix returns an object of class ModelEnv-class - a high level object for storing data improving upon the capabilities of simple data matrices.

Funny things may happen if the inpiut and response matrices do not have distinct column names and the data new data are supplied via the get and set slots.

Value

An object of class ModelEnv-class.

Examples

### use Sepal measurements as input and Petal as response
data(iris)
me <- ModelEnvMatrix(iris[,1:2], iris[,3:4])
me

### extract data from the ModelEnv object
dim(me@get("designMatrix"))
summary(me@get("responseMatrix"))

### subsets and missing values
iris[1,1] <- NA
me  <- ModelEnvMatrix(iris[,1:2], iris[,3:4], subset=1:5, na.action=na.omit)

## First case is not complete, so me contains only cases 2:5
me
me@get("designMatrix")
me@get("responseMatrix")

## use different cases
me@set(data=iris[10:20,])
me@get("designMatrix")

## these two should be the same
stopifnot(all.equal(me@get("responseMatrix"), as.matrix(iris[10:20,3:4])))

Model Predictions

Description

A function for predictions from the results of various model fitting functions.

Usage

Predict(object, ...)

Arguments

Details

A somewhat improved version of predict for models fitted with objects of class StatModel-class.

Value

Should return a vector of the same type as the response variable specified for fitting object.

Examples

df <- data.frame(x = runif(10), y = rnorm(10))
mf <- dpp(linearModel, y ~ x, data = df)
Predict(fit(linearModel, mf))

Class "StatModel"

Description

A class for unfitted statistical models.

Objects from the Class

Objects can be created by calls of the form new("StatModel", ...).

Slots

name:

Object of class "character", the name of the model.

dpp:

Object of class "function", a function for data preprocessing (usually formula-based).

fit:

Object of class "function", a function for fitting the model to data.

predict:

Object of class "function", a function for computing predictions.

capabilities:

Object of class "StatModelCapabilities".

Methods

fit

signature(model = "StatModel", data = "ModelEnv"): fit model to data.

Details

This is an attempt to provide unified infra-structure for unfitted statistical models. Basically, an unfitted model provides a function for data pre-processing (dpp, think of generating design matrices), a function for fitting the specified model to data (fit), and a function for computing predictions (predict).

Examples for such unfitted models are provided by linearModel and glinearModel which provide interfaces in the "StatModel" framework to lm.fit and glm.fit, respectively. The functions return objects of S3 class "linearModel" (inheriting from "lm") and "glinearModel" (inheriting from "glm"), respectively. Some methods for S3 generics such as predict, fitted, print and model.matrix are provided to make use of the "StatModel" structure. (Similarly, survReg provides an experimental interface to survreg.)

Examples

  ### linear model example
  df <- data.frame(x = runif(10), y = rnorm(10))
  mf <- dpp(linearModel, y ~ x, data = df)
  mylm <- fit(linearModel, mf)

  ### equivalent
  print(mylm)
  lm(y ~ x, data = df)

  ### predictions
  Predict(mylm, newdata =  data.frame(x = runif(10)))

Class "StatModelCapabilities"

Description

A class describing capabilities of a statistical model.

Objects from the Class

Objects can be created by calls of the form new("StatModelCapabilities", ...).

Slots

weights:

Object of class "logical"

subset:

Object of class "logical"

Methods

No methods defined with class "StatModelCapabilities" in the signature.

Get Information on Fitted Objects

Description

Returns descriptive information about fitted objects.

Usage

info(object, which, ...)
## S4 method for signature 'ANY,missing'
info(object, which, ...)
infoCheck(object, which, ...)

Arguments

Details

Function info can be used to access slots of fitted objects in a portable way.

Function infoCheck returns a logical value that is TRUE if the requested information can be computed from the object.

Author(s)

Friedrich Leisch