Common S3 Generics

These are S3 generic functions that can be used to minimize package dependencies across multiple packages that have the same method.

Usage

fit(object, data, params, ...)

tune(model, ...)

get_pos_class(model, ...)

get_model_classes(model, ...)

get_model_features(model, ...)

get_model_coef(model, ...)

get_model_type(model, ...)

get_model_params(model, ...)

calc_model_metrics(model, data, ...)

cross_validate(model, data, k, ...)

calc_predictions(model, newdata, ...)

Arguments

data

The data on which to assess the model.

params

Model parameters. Optional depending on model type.

...

Used for extensibility of downstream S3 methods.

model, object

A model object.

k

The number of folds to perform (k-fold cross-validation).

newdata

The test set (data.frame) containing (features) corresponding to the model parameters. For some models, if newdata = NULL, training (or out-of-bag) predictions are returned.

Functions

• fit(): A S3 generic to "fit" a model.

• tune(): A S3 generic to "tune" a model.

• get_pos_class(): For a binary classification model, determine the positive class, i.e. the "disease" or "case" class.

• get_model_classes(): In a binary classification model, determine the classes in the model.

• get_model_features(): Returns a character vector of the model predictors/covariates contained within a model object. After interrogating the model object, this function returns the covariates (features) contained in the object, both proteomic or clinical meta data.

• get_model_coef(): Returns a named numeric vector of the regression coefficients of a linear model contained within a model object.

• get_model_type(): A S3 generic to determine the model type, e.g. classification, regression, or survival.

• get_model_params(): A S3 generic to determine the model parameters.

• calc_model_metrics(): A S3 generic to calculate performance metrics of a model.

• cross_validate(): A S3 generic to perform k-fold cross-validation for a given model.

• calc_predictions(): Calculate predictions of a given model on a test set defined by newdata. The return object is always a data.frame class object in 1 of 3 formats, depending on the model type, and containing model predictions for newdata. The data frame is named according to:

  Model Type	Format	Example (n = 1)
  continuous	pred_<endpoint>	data.frame(pred_vo2max = 303.9)
  classification	pred_class, prob_<class1>, prob_<class2>	data.frame(pred_class = "nash", prob_normal = 0.2, prob_nash = 0.8)
  survival	pred_status, risk_<status1>, risk_<status2>	data.frame(pred_status = "MI", risk_noMI = 0.2, risk_MI = 0.8)

Note

get_model_coef(): non-linear models will not have a linear predictor, and thus return NULL. See specific S3 method ?get_model_coef.class for details.

calc_predictions(): not all classification models will have a linear predictor, any non-GLM model, e.g. random forest or SVM, will not.

Model Features Classes

Class	Model type	See examples
"glm"	Logistic & linear regression	glm()
"glmnet"	Regularized logistic or linear regression	glmnet::glmnet()
"naiveBayes"	Standard naive Bayes	e1071::naiveBayes()
"fit_nb"	Naive Bayes (robust pars)	fit_nb()
"randomForest"	Random Forests	randomForest::randomForest()
"lda"	Linear Discriminant Analysis	MASS::lda()
"kknn"	k-nearest neighbor	kknn::kknn()
"gbm"	generalized boosted regression models	gbm::gbm()
"svm"	Support Vector Machines	e1071::svm()
"survreg"	Survival models	survival::survreg()
"psm"	Survival models	rms::psm()
"coxnet2"	Regularized cox	fitCoxnet()
"survregnet"	Regularized survival	fitSurvregnet()
"train"	caret models	caret::train()

Author

Stu Field