fs.summary: Feature Ranking Resampling Summary

Description Usage Arguments Value Note Author(s) See Also Examples

View source: R/fs.summary.R

Description

Wrapper that aggregates results obtained from a feature ranking resampling strategy. If p values have been calculated by the feature ranking technique on the overall dataset, adjusted p-values by one of the methods available in p.adjust are also returned.

Usage

1
  fs.summary(res1,res2,padjust="fdr",sorting=TRUE)

Arguments

res1

A list returned from feat.rank.re.

res2

A list returned from fs.mrpval.

padjust

One of the methods in p.adjust.

sorting

Should the results be sorted according to the feature ordering calculated on the overall data?

Value

A matrix of statistics. For details, see Note below.

Note

The output matrix with number of rows corresponding to the number of variables and number of columns to:

Author(s)

David Enot and Wanchang Lin dle, [email protected].

See Also

feat.rank.re, fs.mrpval, p.adjust

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
27
28
29
30
31
32
33
34
35
## load abr1
data(abr1)
y   <- factor(abr1$fact$class)
x <- preproc(abr1$pos , y=y, method=c("log10","TICnorm"),add=1)[,110:500]  
## Select classes 1 and 2
dat <- dat.sel(x, y, choices=c("1","2"))
x <- dat$dat[[1]]
y <- dat$cl[[1]]

## partitioning
pars   <- valipars(sampling="boot",niter=2,nreps=5)
tr.idx <- trainind(y,pars=pars)

## multiple rankings using AUC
z      <- feat.rank.re(x,y,method="fs.auc",pars = pars,tr.idx=tr.idx)

## Compute stability mr-p value using the 75% worst features as irrelevant
res <- fs.mrpval(z,qtl=0.25)

## print content of res
names(res)

res.1 <- fs.summary(z, res, sorting=TRUE)

## Print the 10 best features
print(res.1[1:10,])

##### Example of output with a feature ranking technique that returns p-value
z      <- feat.rank.re(x,y,method="fs.welch",pars = pars,tr.idx=tr.idx)
res <- fs.mrpval(z,qtl=0.25)
names(res)
## p-value correction with fdr
res.1 <- fs.summary(z, res, padjust = "fdr", sorting=TRUE)
## Print the 10 best features
res.1[1:10,]

wilsontom/FIEmspro documentation built on Feb. 19, 2018, 9:03 a.m.