cvCompute: Utility Functions for the cv Package

In cv: Cross-Validating Regression Models

Utility Functions for the cv Package

Description

These functions are primarily useful for writing methods for the cv() generic function. They are used internally in the package and can also be used for extensions (see the vignette "Extending the cv package, vignette("cv-extend", package="cv")).

Usage

cvCompute(
  model,
  data = insight::get_data(model),
  criterion = mse,
  criterion.name,
  k = 10L,
  reps = 1L,
  seed,
  details = k <= 10L,
  confint,
  level = 0.95,
  method = NULL,
  ncores = 1L,
  type = "response",
  start = FALSE,
  f,
  fPara = f,
  locals = list(),
  model.function = NULL,
  model.function.name = NULL,
  ...
)

cvMixed(
  model,
  package,
  data = insight::get_data(model),
  criterion = mse,
  criterion.name,
  k,
  reps = 1L,
  confint,
  level = 0.95,
  seed,
  details,
  ncores = 1L,
  clusterVariables,
  predict.clusters.args = list(object = model, newdata = data),
  predict.cases.args = list(object = model, newdata = data),
  fixed.effects,
  ...
)

cvSelect(
  procedure,
  data,
  criterion = mse,
  criterion.name,
  model,
  y.expression,
  k = 10L,
  confint = n >= 400,
  level = 0.95,
  reps = 1L,
  save.coef,
  details = k <= 10L,
  save.model = FALSE,
  seed,
  ncores = 1L,
  ...
)

folds(n, k)

fold(folds, i, ...)

## S3 method for class 'folds'
fold(folds, i, ...)

## S3 method for class 'folds'
print(x, ...)

GetResponse(model, ...)

## Default S3 method:
GetResponse(model, ...)

## S3 method for class 'merMod'
GetResponse(model, ...)

## S3 method for class 'lme'
GetResponse(model, ...)

## S3 method for class 'glmmTMB'
GetResponse(model, ...)

## S3 method for class 'modList'
GetResponse(model, ...)

Arguments

model	a regression model object.
data	data frame to which the model was fit (not usually necessary, except for cvSelect()).
criterion	cross-validation criterion ("cost" or lack-of-fit) function of form f(y, yhat) where y is the observed values of the response and yhat the predicted values; the default is mse (the mean-squared error).
criterion.name	a character string giving the name of the CV criterion function in the returned "cv" object).
k	perform k-fold cross-validation (default is 10); k may be a number or "loo" or "n" for n-fold (leave-one-out) cross-validation; for folds(), k must be a number.
reps	number of times to replicate k-fold CV (default is 1).
seed	for R's random number generator; optional, if not supplied a random seed will be selected and saved; not needed for n-fold cross-validation.
details	if TRUE (the default if the number of folds k <= 10), save detailed information about the value of the CV criterion for the cases in each fold and the regression coefficients with that fold deleted.
confint	if TRUE (the default if the number of cases is 400 or greater), compute a confidence interval for the bias-corrected CV criterion, if the criterion is the average of casewise components.
level	confidence level (default 0.95).
method	computational method to apply; use by some cv() methods.
ncores	number of cores to use for parallel computations (default is 1, i.e., computations aren't done in parallel).
type	used by some cv() methods, such as the default method, where type is passed to the type argument of predict(); the default is type="response", which is appropriate, e.g., for a "glm" model and may be recognized or ignored by predict() methods for other model classes.
start	used by some cv() methods; if TRUE (the default is FALSE), the start argument, set to the vector of regression coefficients for the model fit to the full data, is passed to update(), possibly making the CV updates faster, e.g. for a GLM.
f	function to be called by cvCompute() for each fold.
fPara	function to be called by cvCompute() for each fold using parallel computation.
locals	a named list of objects that are required in the local environment of cvCompute() for f() or fPara().
model.function	a regression function, typically for a new cv() method, residing in a package that's not a declared dependency of the cv package, e.g., nnet::multinom.
model.function.name	the quoted name of the regression function, e.g., "multinom".
...	to match generic; passed to predict() for the default method, and to fPara() (for parallel computations) in cvCompute().
package	the name of the package in which mixed-modeling function (or functions) employed resides; used to get the namespace of the package.
clusterVariables	a character vector of names of the variables defining clusters for a mixed model with nested or crossed random effects; if missing, cross-validation is performed for individual cases rather than for clusters
predict.clusters.args	a list of arguments to be used to predict the whole data set from a mixed model when performing CV on clusters; the first two elements should be model and newdata; see the "Extending the cv package" vignette (vignette("cv-extend", package="cv")).
predict.cases.args	a list of arguments to be used to predict the whole data set from a mixed model when performing CV on cases; the first two elements should be model and newdata; see the "Extending the cv package" vignette (vignette("cv-extend", package="cv")).
fixed.effects	a function to be used to compute fixed-effect coefficients for cluster-based CV when details = TRUE.
procedure	a model-selection procedure function (see Details).
y.expression	normally the response variable is found from the model argument; but if, for a particular selection procedure, the model argument is absent, or if the response can't be inferred from the model, the response can be specified by an expression, such as expression(log(income)), to be evaluated within the data set provided by the data argument.
save.coef	save the coefficients from the selected models? Deprecated in favor of the details argument; if specified, details is set is set to the value of save.coef.
save.model	save the model that's selected using the full data set.
n	number of cases, for constructed folds.
folds	an object of class "folds".
i	a fold number for an object of class "folds".
x	a "cv", "cvList", or "folds" object to be printed

Value

The utility functions return various kinds of objects:

• cvCompute() returns an object of class "cv", with the CV criterion ("CV crit"), the bias-adjusted CV criterion ("adj CV crit"), the criterion for the model applied to the full data ("full crit"), the confidence interval and level for the bias-adjusted CV criterion ("confint"), the number of folds ("k"), and the seed for R's random-number generator ("seed"). If details=TRUE, then the returned object will also include a "details" component, which is a list of two elements: "criterion", containing the CV criterion computed for the cases in each fold; and "coefficients", regression coefficients computed for the model with each fold deleted. Some cv() methods calling cvCompute() may return a subset of these components and may add additional information. If reps > 1, then an object of class "cvList" is returned, which is literally a list of "cv" objects.

• cvMixed() also returns an object of class "cv" or "cvList".

• cvSelect returns an object of class "cvSelect" inheriting from "cv", or an object of class "cvSelectList" inheriting from "cvList".

• folds() returns an object of class folds, for which there are fold() and print() methods.

• GetResponse() returns the (numeric) response variable from the model.

  The supplied default method returns the model$y component of the model object, or, if model is an S4 object, the result returned by the get_response() function in the insight package. If this result is NULL, the result of model.response(model.frame(model)) is returned, checking in any case whether the result is a numeric vector.

  There are also "lme", "merMod" and "glmmTMB" methods that convert factor responses to numeric 0/1 responses, as would be appropriate for a generalized linear mixed model with a binary response.

Functions

• cvCompute(): used internally by cv() methods (not for direct use); exported to support new cv() methods.

• cvMixed(): used internally by cv() methods for mixed-effect models (not for direct use); exported to support new cv() methods.

• cvSelect(): used internally by cv() methods for cross-validating a model-selection procedure; may also be called directly for this purpose, but use via cv() is preferred. cvSelect() is exported primarily to support new model-selection procedures.

• folds(): used internally by cv() methods (not for direct use).

• fold(): to extract a fold from a "folds" object.

• fold(folds): fold() method for "folds" objects.

• print(folds): print() method for "folds" objects.

• GetResponse(): function to return the response variable from a regression model.

• GetResponse(default): default method.

• GetResponse(merMod): "merMod" method.

• GetResponse(lme): "lme" method.

• GetResponse(glmmTMB): "glmmTMB" method.

• GetResponse(modList): "modList" method.

See Also

cv, cv.merMod, cv.function.

Examples

fit <- lm(mpg ~ gear, mtcars)
GetResponse(fit)

set.seed(123)
(ffs <- folds(n=22, k=5))
fold(ffs, 2)

cv documentation built on Sept. 23, 2024, 1:08 a.m.