summ.ftrank: Summarise multiple resampling based feature ranking outputs

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

View source: R/ftrank.R

Description

This routine performs further computations on a list of feat.rank.re objects contained in a mfr.obj object (see ftrank.agg). If only one feat.rank.re result from feat.rank.re is analyzed, it is easier to pass it first to ftrank.agg (see example).Two calculations are made: 1) Computation of the pseudo mrp-value from the resampling based feature (see fs.mrpval) and 2) adjusted p values if p values have calculated (see p.adjust).

Usage

1
summ.ftrank(lclas, lmod = NULL, qtl = 0.25, padjust = "fdr")

Arguments

lclas

mfr.obj object - See details in ftrank.agg

lmod

List of models to be considered in lclas

qtl

Quantile - See details in fs.mrpval

padjust

p value adjustement method - See details in p.adjust

Details

The resulting list as two component: one is equal to the total number of feat.rank.re objects (i.e one resampling experiment) and one field is a table that summarises each feat.rank.re (as for ftrank.agg). In the first component, each table may have a different number of columns depending if the FR method outputs p-values or not:

Stat:

Original statistics calculated on the overall data.

Rank:

Original feature rank calculated on the overall data.

pval:

Original feature p-value calculated on the overall data (optional).

pval-xxx:

Feature adjusted p-value by method xxx if p-value available (i.e. previous column).

mrpval-xxx:

Pseudo multiple resampling p-value using a given qtl value xxx.

AvgRk:

Average feature rank calculated from the ranks found for each training data partition.

SdevRk:

Associated feature rank standard deviation calculated from the ranks found for each training data partition.

Value

mfr.sum object:

frsum

List of tables corresponding to each feat.rank.re object - See details

frdef

Summary of each feat.rank.re object as in ftrank.agg

Author(s)

David Enot [email protected]

See Also

fs.mrpval, tidy.ftrank, p.adjust

Examples

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
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.sel1(x, y, pwise="1",mclass=NULL,pars=valipars(sampling="boot",niter=2,nreps=5))

reswelch = feat.rank.re(dat[[1]],method="fs.welch")
mfr=ftrank.agg(reswelch)
print(mfr$frdef)
frsum=summ.ftrank(mfr,lmod=1,qtl=.3)

## print the FR components
print(frsum$frdef)

## have a look at the first 5 variables
print(frsum$frsum[[1]][1:5,])

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