pairwise.model.stability {OmicsMarkeR} | R Documentation |
Conducts all pairwise comparisons of each model's selected features selected following bootstrapping. Also known as the function perturbation ensemble approach
pairwise.model.stability(features, stability.metric, nc)
features |
A matrix of selected features |
stability.metric |
string indicating the type of stability metric.
Avialable options are |
nc |
Number of original features |
A list is returned containing:
comparisons |
Matrix of pairwise comparisons |
overall |
The average of all pairwise comparisons |
Charles Determan Jr
He. Z. & Weichuan Y. (2010) Stable feature selection for biomarker discovery. Computational Biology and Chemistry 34 215-225.
# 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)