fs.mrpval: Significance of Feature Ranking

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

Description

Computation of the pseudo mrp-value from a resampling based feature ranking strategy. qtl represents the fraction of presumedly informative features. The decision is based on the average rank across all resampling steps. 1-qtl represents the fraction of features that serves to estimate the null distribution of ranks (i.e. ranks of uninformative variables).

Usage

1
  fs.mrpval(x,qtl=0.75)

Arguments

x

A list returned from feat.rank.re.

qtl

A numeric value of probability with values in [0,1].

Value

A list with components:

stats

Original feature ranking statistics.

fs.rank

Feature ranking vector.

fs.order

Feature order vector.

sdrank

Feature rank standard deviation.

mrpval

Individual feature mrp-value.

Ug

Uninformative variables.

nnull

Total number of uninformative variables.

qtl

Quantile qtl used.

Author(s)

David Enot [email protected] and Wanchang Lin [email protected]

References

Zhang, C., Lu,X. and Zhang, X. (2006). Significance of Gene Ranking for Classification of Microarray Samples. IEEE/ACM Transactions on Computational Biology and Bioinformatics, VOL. 3, NO. 3, pp. 312-320.

See Also

feat.rank.re, fs.summary

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
## 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 25% worst features as irrelevant
res <- fs.mrpval(z,qtl=0.75)

## print content of res
names(res)

## list of features to form the null distribution of ranks
print(res$Ug)

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