fs.summary: Feature Ranking Resampling Summary

In aberHRML/FIEmspro: Flow In-jection Electrospray Mass Spectrometry Processing: \\ data processing, classification modelling and variable selection in metabolite fingerprinting

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

  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:

• Feature ranking quantity computed on the whole dataset

• Feature rank in decreasing order of saliency

• p-value if available with feature ranking technique (optional)

• adjusted p-value if p-value available (optional)

• average feature rank across every resampling steps

• standard deviation of the feature rank across every resampling steps

• pseudo p-value calculated from the resampling strategy

Author(s)

David Enot and Wanchang Lin dle, [e-mail].

See Also

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

Examples

## 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,]

aberHRML/FIEmspro documentation built on May 16, 2019, 6:56 p.m.