Customized boxplot visualization of benchmark results

This is a convenience function to create customized boxplots for specific benchmark criteria such as runtime, statistical significance and phenotype relevance.

Usage

bpPlot(data, what = c("runtime", "sig.sets", "rel.sets", "typeI"))

Arguments

data

Numeric matrix or list of numeric vectors. In case of a matrix, column names are assumed to be method names and rownames are assumed to be dataset IDs. In case of a list, names are assumed to be method names and each element corresponds to a numeric vector with names assumed to be dataset IDs.

what

Character. Determines how the plot is customized. One of

• runtime: displays runtime of methods across datasets,

• sig.sets: displays percentage of significant gene sets,

• rel.sets: displays phenotype relevance scores,

• typeI: displays type I error rates.

Value

None. Plots to a graphics device.

See also

evalNrSigSets to evaluate fractions of significant gene sets; evalRelevance to evaluate phenotype relevance of gene set rankings.

Author

Ludwig Geistlinger <Ludwig.Geistlinger@sph.cuny.edu>

Examples

    # simulated setup:
    # 3 methods & 5 datasets
    methods <- paste0("m", 1:3)
    data.ids <- paste0("d", 1:5)

    # runtime data
    rt <- vapply(1:3, function(m) runif(5, min = m, max = m+1), numeric(5))
    rownames(rt) <- data.ids 
    colnames(rt) <- methods

    # plot
    bpPlot(rt, what = "runtime")