mc.agg: Aggregation of classification results

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

Description

Aggregate accest objects and list of accest objects to form mc.agg object. The main utilities of this function is to concatenate in a single list various results derived from several accest calls in order to facilitate post analysis additional treatments as well as exporting the results.

Usage

mc.agg(...)

## Default S3 method:
mc.agg(...)

Arguments

...	accest objects and or list of accest objects

Details

The length of the resulting list is equal to the total number of accest objects (i.e one resampling experiment) plus one field that summarises each accest. The later is a table with 8 columns which are automatically generated to and also avoid confusions if the same method is applied on the same discrimination problem but with different settings, different resampling partitioning or even different data sets. Note that columns 1 to 5 are automatically generated from the output of accest where as columns 6 is based on columns 2-4. Each column is described as follows:

Mod:

Unique identifier for each resampling based feature rankings.

Alg:

Name of the classification technique as specified in the call of accest.

Arg:

Arguments passed to the classifier during the call of accest.

Pars:

Summary of the resampling strategy adopted during the call of accest.

Dis:

Discrimination task involved. By default, this is equal to the actual levels of the class vector passed to accest separated by ~.

AlgId:

Unique algorithm identifier based on the columns Alg, Arg and Pars so that no confusion is possible with Alg column if several classification have been performed with the same discrimination technique but with different parameters and/or resampling strategy. This column can be modified by the user.

DisId:

Unique algorithm identifier based on the columns Dis in order to simplified the name of the discrimination task if there are many classes involved and/or class level have a long name. This column can be modified by the user.

Other:

Empty column that can be amended to store extra information.

Value

mc.agg objects:

clas	List of accest objects
cldef	Summary of each accest object - See details

Author(s)

David Enot [e-mail]

See Also

accest,mc.summary, mc.comp.1, mc.roc

Examples

data(iris)
dat=as.matrix(iris[,1:4])
cl=as.factor(iris[,5])
lrnd=sample(1:150)[1:50]
cl[lrnd]=sample(cl[lrnd])
pars   <- valipars(sampling = "boot",niter = 2, nreps=10)
dat1=dat.sel1(dat,cl,pwise=list(),mclass=list(),pars=pars)

res1=accest(dat1[[1]],clmeth="randomForest",ntree=100,seed=1)
res2=lapply(dat1,function(x) accest(x,clmeth="lda"))

mc=mc.agg(res1,res2)
###Print the content
mc

### Classification task num. 1:
mc$clas[[1]]

## As other functions are using the column 5 and 6 to sort and print the results,
## you can replace them by something more informative
## for e.g. Alg and Dis
mc$cldef[,6]<-mc$cldef[,2]
mc$cldef[,7]<-c("set~vir","set~vir","set~vir","ver~vir","ver~vir")
mc$cldef

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