clsMRes: Perform Clustering for Multiple Resolution Parameters

In ornithos/nectr: Exploring Datasets via Clustering with TURN-RES.

Description

The initial stage of cluster investigation using nectr. clsMRes shows how the clusters grow and agglomerate across increasing resolutions of the clsTurnRes algorithm.

Usage

clsMRes(data, r.start = NA, r.max = Inf, keep = FALSE, ...)

Arguments

data	Required. The data to be clustered. Each column must be a dimension; no labels are permitted. data must be of class data.frame.
keep	Optional. Whether to keep the clustering vector associated with each clsTurnRes iteration. By default the algorithm keeps the clustering vector found on each iteration, but one may wish to discard them if the vectors are large.
r.start	Optional. For user specified resolution step sequence. This specifies the first number in the sequence.
r.max	Optional. For user specified resolution step sequence. This specifies the last number in the sequence - ie the search is terminated if the parameter value is not less than r.max.
...	Optional (additional arguments to be passed to clsTurnRes)

Details

Automatic function - should be no need for parameter choice if uncertain. The first goal is to find an appropriate sequence of resolutions to run through the clsTurnRes algorithm. A starting position is determined automatically, but it may be quicker to specify the starting resolution once the analyst is familiar with the dataset.

Value

An object of class clsMR. Primary output is to demonstrate structure through increasing resolution parameter. Standard visualisation available through the plot generic function. Object can be interrogated by cluster via the subset [] operator to give details of resolution used.

Note

In order to avoid copying the dataset to each clsMR object, instead the name is saved as item $dataset.name. The data will then be retrieved in function calls via get(dataset.name, env = .GlobalEnv), which means that the user must ensure that the dataset variable name is not changed. There are obvious problems with this approach but the alternative is to waste memory copying potentially large datasets every time the function is called.

Author(s)

Alex Bird, [e-mail]

See Also

clsTurnRes, the clustering algorithm used.

Examples

x = rnorm(30,0,1);
mean(x) #approx 0.5. Nothing to do with this!

ornithos/nectr documentation built on May 24, 2019, 3:57 p.m.