pairwise.model.stability: Pairwise Model Stability Metrics

In OmicsMarkeR: Classification and Feature Selection for 'Omics' Datasets

Description

Conducts all pairwise comparisons of each model's selected features selected following bootstrapping. Also known as the function perturbation ensemble approach

Usage

pairwise.model.stability(features, stability.metric, nc)

Arguments

features	A matrix of selected features
stability.metric	string indicating the type of stability metric. Avialable options are "jaccard" (Jaccard Index/Tanimoto Distance), "sorensen" (Dice-Sorensen's Index), "ochiai" (Ochiai's Index), "pof" (Percent of Overlapping Features), "kuncheva" (Kuncheva's Stability Measures), "spearman" (Spearman Rank Correlation), and "canberra" (Canberra Distance)
nc	Number of original features

Value

A list is returned containing:

comparisons	Matrix of pairwise comparisons
overall	The average of all pairwise comparisons

Author(s)

Charles Determan Jr

References

He. Z. & Weichuan Y. (2010) Stable feature selection for biomarker discovery. Computational Biology and Chemistry 34 215-225.

See Also

pairwise.stability

Examples

# pairwise.model.stability demo
# For demonstration purposes only!!!
some.numbers <- seq(20)

# A list containing the metabolite matrices for each algorithm
# As an example, let's say we have the output from two different models
# such as plsda and random forest.
# matrix of Metabolites identified (e.g. 5 trials)
plsda <- 
    replicate(5, paste("Metabolite", sample(some.numbers, 10), sep="_"))
rf <-
    replicate(5, paste("Metabolite", sample(some.numbers, 10), sep="_"))

features <- list(plsda=plsda, rf=rf)

# nc may be omitted unless using kuncheva
pairwise.model.stability(features, "kuncheva", nc=20)

Example output

$comparisons
            Resample.1 Resample.2 Resample.3 Resample.4 Resample.5
plsda.vs.rf        0.5        0.5        0.6        0.7        0.4

$overall
[1] 0.54

OmicsMarkeR documentation built on April 28, 2020, 6:54 p.m.