Err: Generic function for computing error rates

In sglOptim: Sparse group lasso generic optimizer

Description

Compute and returns an error rate for each model contained in x. See details for generic use cases.

Usage

Err(object, data, response, ...)

Arguments

object	an object
data	a data object
response	a response object
...	additional parameters (optional)

Details

The following generic use case should be supported (see for example msgl package for an implementation):

1. With fit a sgl fit object with models estimated using x data, the code

   Err(fit, x)

   should return a vector with the training errors of the models.

2. With x.new a new data set with known responses response.new, the code

   Err(fit, x.new, response.new)

   should return a vector with the errors of the models when applied to the new data set.

3. With fit.cv a sgl cross validation object, the code

   Err(fit.cv)

   should return a vector with estimates of the expected generalization errors of the models (i.e. the cross validation errors).

4. If subsampling is supported then, with fit.sub a sgl subsampling object, the code

   Err(fit.sub)

   should return a matrix with the test errors (each column corresponding to a model, i.e. rows corresponds to tests).

Value

a vector of length nmod(object) or a matrix with nmod(object) columns containing error rates for the models

Author(s)

Martin Vincent

See Also

compute_error

sglOptim documentation built on May 2, 2019, 5:55 p.m.