bsPAMR: Perform PAMR bootstrapping.

In bootfs: Derive Robust Feature Sets for Two or Multiclass Classification Problems

Description

Use PAMR algorithm and bootstrapping to derive a robust feature set.

Usage

bsPAMR(logX, groupings, DIR=NULL, params = NULL)

Arguments

logX	The data matrix.
groupings	The grouping variable.
DIR	The output base directory.
params	A parameter list with the following elements: seed: A random seed. bstr: Number of bootstrap iterations. jitter: Boolean. Use some additional noise for constant values. ncv: Number of inner crossvalidation folds. max_allowed_feat: Maximum number of allowed features in the feature set. n.threshold: Threshold parameter for PAMR. saveres: Boolean. Stored results of bootstrapping in an R-data object.

Details

Usually called from doBS.

Value

The result list of the bootstrapping, which is a list: For each classification instance in groupings, one element is saved, each a list with bstr elements, from which each again is a list holding the following elements:

histtr	training object generated with pamr.train.
histcv	cv object generated with pamr.cv
tmin	Minimal threshold value for the feature selection.
selected	Indices of selected features at tmin.
selected_names	Names of selected features at tmin.

To get to one bootstrapping element, one could use for example: returnobject[[1]][[1]][["tmin"]] to get access to the minimal threshold of the first bootstrapping iteration in the first classification problem.

Author(s)

Christian Bender ([e-mail])

See Also

doBS

Examples

## Not run: 
#todo

## End(Not run)

bootfs documentation built on May 2, 2019, 5:50 p.m.