Summarise the results of a bootstrapping analysis.

Share:

Description

Get frequencies of features and the importance graph by combining all bootstrapping runs and classification methods.

Usage

1
2
resultBS(results, DIR=NULL, vlabel.cex = 3, filter = 10, 
		useresults=1:length(results), plot=TRUE)

Arguments

results

Result object from doBS.

DIR

Output base directory. If not NULL, output is stored in this folder. Creates the following files: importance_graph_combined.pdf, adj_sorted.csv and tophits.csv.

vlabel.cex

Node label expansion factor.

filter

Edge occurrence filter. Only show edges with larger occurrence count than filter.

useresults

Vector of integers. Specify the elements of the results list to use for summarisation. Defaults to all (i.e. use results from all feature selection algorithms).

plot

Boolean. If TRUE, show the importance graph plot.

Details

Derives the ranking of features across all bootstrapping algorithms, according to how often they occur in all bootstrapping runs. resultBS is a list with one element for each classification method. In each of these elements, for every separate classification task (e.g. different outcome groups) one sub list is stored. resultBS extracts each individual classification task and makes the importance graph and adjacency matrix as described below.

For each classification task, the intersection of the selected features across all algorithms in each bootstrapping iteration is created. Occurences of the intersected features are counted to create the final ranking.

Value

A list with one element for each classification task (i.e. group classification). Each element holds the following sub elements:

allsignatures

Intersection feature sets for each bootstrap run.

allprots

Union of all intersection features.

adj

Adjacency matrix holding the occurence counts.

tophits

Top selected features, ranked according to there occurrence.

Author(s)

Christian Bender

See Also

doBS

Examples

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
## Not run: 
# library(bootfs)
set.seed(1234)
data <- simDataSet(nsam=30, ngen=100, sigma=2, plot=TRUE)
logX <- data$logX
groupings <- data$groupings

params <- control_params(bstr=5, maxiter=100, maxevals=50, maxRuns=30)
	
## run the bootstrapping
retBS <- doBS(logX, groupings, 
	fs.methods=c("pamr","scad","rf_boruta"),
	DIR=NULL, 
	params=params)

## create the importance graph for all methods
## and export the adjacency matrix containing the 
## numbers of occuerrences of the features, as well 
## as the top hits.
res <- resultBS(retBS, DIR=NULL, vlabel.cex = 3, filter = 0, saveres = FALSE)

## remove the created directory
system("rm -rf bs")


## End(Not run)

Want to suggest features or report bugs for rdrr.io? Use the GitHub issue tracker.