archi: Archipelago Analysis
In optpart: Optimal Partitioning of Similarity Relations

Description Usage Arguments Details Value Author(s) Examples

Archipelago analysis finds connected clusters in a dissimilarity matrix. Samples in the same cluster are at most alpha dissimilar to at least one other sample in the cluster, and are more than alpha dissimilar to all samples in all other clusters. The solution is equivalent to slicing a nearest neighbor cluster analysis at alpha, but does not require (or produce) a hierarchical structure.

1	archi(dist,alpha)

`dist`	an object of class ‘dist’ from `dist`, `vegdist`, or `dsvdis`
`alpha`	the dissimilarity threshold to establish the relationship

Archipelago analysis is a topological, as opposed to metric space, cluster routine that returns connected clusters. Every sample in a cluster is connected by a path with step lengths of at most alpha dissimilarity to every other sample in the cluster, and is more than alpha dissimilar to all other samples in all other clusters.

produces an object of class ‘clustering’, a list with a vector ‘clustering’ of cluster memberships

David W. Roberts droberts@montana.edu

data(shoshveg) # produces a vegetation dataframe
dis.bc <- dsvdis(shoshveg,'bray/curtis') 
               # produces a Bray/Curtis dissimilarity matrix
arc.50 <- archi(dis.bc,0.5) # clusters at 0.5 dissimilarity 
table(arc.50$clustering)