| bestMclust {edci} | R Documentation |
Chooses the 'best' regression cluster(s), if the number of true clusters is known.
bestMclust(clust, nc=1, crit="value") projMclust(clust, x, y) envMclust(clust, x, y, dist=0)
clust |
Cluster object returned by oregMclust or circMclust. |
nc |
Number of 'best' clusters. |
crit |
Name of the column to determine the best clusters. |
x,y |
Original observations. |
dist |
Maximal distance of observation from cluster center. |
oregMclust and circMclust return a matrix containing not
only the parameters of the found clusters but the value of the heights
of the corresponding local maxima as well as how often each cluster is
found. Both are reasonable criteria for choosing 'best' clusters, which
can be done by bestMclust.
Additional criteria could be the number of observations projected to
each cluster or the number of observations lying in a certain
neighbourhood of the cluster center point.
projMclust adds a column proj to clust which
contains the number of points belonging to each cluster in the sense
that each observation belongs to the cluster with shortest orthogonal
distance. If clust is comming from circMclust a second
column projrel is added which contains this number relativ to
the radius of the particular circle.
envMclust adds a column env to clust which
contains the number of observations lying in a
dist-neighbourhood of each cluster center. If clust is
comming from circMclust a second column envrel is added
which contains this number relativ to the radius of the particular
circle.
Both functions return a matrix of clusters.
Tim Garlipp, garlipp@mathematik.uni-oldenburg.de
Müller, C.H., Garlipp, T. (2003) Simple consistent cluster methods based on redescending M-estimators with an application to edge identification in images, to appear in JMVA.