The level set estimation and visualization.
- Paper reference: Chen, Yen-Chi, Christopher R. Genovese, and Larry Wasserman. "Density Level Sets: Asymptotics, Inference, and Visualization." arXiv preprint arXiv:1504.05438 (2015).
- Contact: yenchic@uw.edu
HDLV_RS.Ris for Rstudio that allows a visualzation using the Shiny app.HDLV.Ris for ordinary R.Example.Ris an example running the script.
HDLV.default = function(data,h = NULL, lv_seq=NULL, lv_min=NULL, eps=1.0e-8, max.iterations=1000, cut=0.1, ...)
- The main function for computing level set estimation and visualization.
- Inputs:
- data: Input data matrix.
- h: Smoothing parameter. Default NULL will choose from the normal reference rule.
- lv_seq: A sequence of density levels. Default NULL will choose from a sequence of level.
- lv_min: Minimal density level (to stablized the level sets). Default NULL will pick 0.05*p_max.
- eps: The tolerance. If mean shift moves less than this value, we will consider it done.
- max.iterations: Maximal number of iteration for mean shift.
- cut: The cut for heirachical clustering (we cut the dedrogram by height = cut*h).
- Outputs:
- An S4 object "HDLV" consisting of the following attributes:
- data: The original data.
- h: The smoothing parameter.
- lv.min: The minimal level.
- density: The density by kernel density estimator.
- MS.labels: The cluster labels from mode clustering.
- modes: The local modes corresopnding to each cluster.
- sig.clu: The significant clusters (local modes).
- lv_seq: The sequence of level sets.
- size: The matrix for the sizes of each significant clusters at different density levels.
- conn: The list of connectivity matrices at each density level.
- An S4 object "HDLV" consisting of the following attributes: