get_what.logreg_screenr: An S3 Method for Extraction of Components from...

In sgutreuter/screenr: Construction of Binary Test-Screening Rules

An S3 Method for Extraction of Components from logreg_screenr Objects

Description

get_what.logreg_screenr extracts components from logreg_screenr-class objects.

Usage

## S3 method for class 'logreg_screenr'
get_what(
  from = NULL,
  what = c("ModelFit", "ROCci", "cvROC", "isROC"),
  ...,
  conf_level = 0.95,
  bootreps = 4000,
  se_min = 0.8
)

Arguments

from	the logreg_screenr-class object from which to extract the component.
what	the (character) name of the component to extract. Valid values are ⁠"ModelFit"⁠, ⁠"ROCci"⁠, ⁠"cvROC"⁠ and ⁠"isROC"⁠.
...	optional arguments to get_what methods.
conf_level	(optional) confidence level for what = ⁠"ROCci"⁠. Default: 0.95.
bootreps	the number of bootstrap replications for estimation of confidence intervals for what = ⁠"ROCci"⁠. Default: 4000.
se_min	minimum value of sensitivity printed for what = ⁠ROCci⁠. Default: 0.8.

Details

get_what is provided to enable easy extraction of components for those who wish to perform computations that are not provided by the coef, plot, predict, print or summary methods.

The following values of what return:

⁠"ModelFit"⁠

the entire glm-class object produced by by glm.

⁠ROCci⁠

a data frame containing cross-validated sensitivities, specificities and their confidence limits, and thresholds.

⁠"cvROC"⁠

the roc-class object produced by roc containing the k-fold cross-validated receiver-operating characteristic.

⁠"isROC"⁠

the roc-class object produced by roc containing the in-sample (overly optimistic) receiver-operating characteristic.

Value

get_what.logreg_screenr returns (invisibly) the object specified by what.

Examples

## Not run: 
attach(uniobj2)
## Get and print cross-validated sensitivities and specificities at
##   thresholds for the local maxima of the ROC curve
myROCci <- get_what(from = uniobj2, what = "ROCci")
print(myROCci)

## End(Not run)

sgutreuter/screenr documentation built on Oct. 19, 2024, 12:49 p.m.