COMMUNAL (version 1.1.0)

Measure Cluster Monotonicty: Measure cluster monotonicty within the output from clusterRange.

Description

These three functions serve to automatically select non-correlated, non-monotonic measures from the output of a run of clusterRange. The user is only required to run getNonCorrNonMonoMeasures; natively, this will return 'numMeasures' measures, selected from all measures that are at or below the mean percentage monotonicity.

The functions measuresCorr and monotoneClusterRange allow for other selection methods by advanced users.

Usage

monotoneClusterRange(clusRange, goodAlgs = NULL)
measuresCorr(clusRange, goodMeasures)
getNonCorrNonMonoMeasures(clusRange, goodMeasures = "all", goodAlgs = NULL, numMeasures = 4)

Arguments

clusRange

The output from a call to clusterRange.

goodMeasures

A subset of validity measures pre-selected by the user. Defaults to "all".

goodAlgs

A subset of clustering algorithms pre-selected by the user. The defauly (NULL) will use all algorithms present in the clusRange object.

numMeasures

The number of clusters to be returned from the clustering of correlations. Literally, call 'cutree' at k=numMeasures on the hierarchical clustering of the non-monotonic measures.

Value

monotoneClusterRange outputs a named numeric vector of the percent monotonicity of all validity measures in clusRange.measuresCorr outputs a correlation matrix of all goodMeasures in clusRange.getNonCorrNonMonoMeasures outputs a character vector of validation measures, suitable for passing into plotRange3D

Details

Advanced users may wish to select validation measures by a guideline of their choosing; these functions can assist in this process as well. In particular, the COMMUNAL manuscript shows a plot of correlations of measures; this was made using the corrplot::corrplot() function on the output from measuresCorr. Such visualization can be helpful in selecting numMeasures.

References

See COMMUNAL manuscript, Sweeney et al., 2015.

See Also

getGoodAlgs

Examples

## output from running \code{clusterRange} on data(BRCA.100)
data(BRCA.results)

monotoneClusterRange(BRCA.results)

measuresCorr(BRCA.results, goodMeasures=c("avg.silwidth", "g3", "sindex"))

getNonCorrNonMonoMeasures(BRCA.results)