dat.sel1: Generate Data Set List

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

Description

Generate subset data sets based on class labels. This function allows generation of a selection of pairwise problems and/or multiple class problems. The main objective of this function is to alleviate generation of several variables containing data matrices, class information or validation strategy. Each subset can enter accest and feat.rank.re analysis without specifying both class and data matrix. Additionally, each subset can also contain validation parameters so that direct comparison between classifiers and feature ranking technique can be easily done.

Usage

  dat.sel1(dat, cl, pwise = NULL, mclass = list(),pars=NULL)

Arguments

dat	A data frame or matrix.
cl	A factor or vector of class.
pwise	The vector or list of class labels to be chosen for binary classification.
mclass	The vector or list of class labels to be chosen to be included in a multiple classification task.
pars	Partitioning information.

Details

This function is used to provide the data set for the binary combination of the class factor. If pwise is list(), the binary combination of for all class labels will be done. If pwise has one class label, the comparisons between this one and any other class are done. If pwise has more than three class lables, enumerate the combinations or permutations of the elements of pwise. For details, see examples below.

Value

A list with components:

nam	Discrimination task.
dat	Subset of the dataset.
cl	Class labels.
pars	Object of type pars.
tr.idx	Object of type trainind.
lsamp	Sample ids in the original data matrix.

Author(s)

David [e-mail]

Examples

data(abr1)
x<-abr1$pos[,110:120]
y<-factor(abr1$fact$day) 

## generate data set with all pairwise containing class "1"
dat1 <- dat.sel1(x,y,pwise="1",mclass=NULL)
unlist(lapply(dat1,function(x) x$name))

## generate data set with pairwise between classes "1" and "H"
dat1 <- dat.sel1(x,y,pwise=c("1","H"),mclass=NULL)
unlist(lapply(dat1,function(x) x$name))

## generate data set with pairwise between classes "1", "2", "H"
dat1 <- dat.sel1(x,y,pwise=c("2","1","H"),mclass=NULL)
unlist(lapply(dat1,function(x) x$name))

## generate data set with all pairwises containing "1" and "H"
dat1 <- dat.sel1(x,y,pwise=list("1","H"),mclass=NULL)
unlist(lapply(dat1,function(x) x$name))

## generate data set with 3 classes "1", "2" and "H"
dat1 <- dat.sel1(x,y,pwise=NULL,mclass=c("1","2","H"))
unlist(lapply(dat1,function(x) x$name))

## generate data set with all pairwises containing "1" and "H"
## on which partitioning is a 1*5CV
pars=valipars(sampling = "cv", niter = 1, nreps = 5)
dat1 <- dat.sel1(x,y,pwise=list("1","H"),mclass=NULL, pars=pars)
unlist(lapply(dat1,function(x) x$name))

print(dat1[[1]])

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