Evaluate Generalized Horizon Labels

Data-driven evaluation of generalized horizon labels using nMDS and silhouette width.

Usage

evalGenHZ(
  obj,
  genhz = GHL(obj, required = TRUE),
  vars,
  non.matching.code = "not-used",
  stand = TRUE,
  metric = "euclidean"
)

Arguments

obj: a SoilProfileCollection object
genhz: name of horizon-level attribute containing generalized horizon labels
vars: character vector of horizon-level attributes to include in the evaluation
non.matching.code: code used to represent horizons not assigned a generalized horizon label
stand: standardize variables before computing distance matrix, passed to cluster::daisy()
metric: distance metric, passed to cluster::daisy()

Value

a list is returned containing:

horizons: c('mds.1', mds.2', 'sil.width', 'neighbor')
stats: mean and standard deviation vars, computed by generalized horizon label
dist: the distance matrix as passed to stats::cmdscale()

Details

Classic multidimensional scaling is performed via stats::cmdscale(). The input distance matrix is generated by cluster::daisy() using (complete cases of) horizon-level attributes from obj as named in vars.

Silhouette widths are computed via cluster::silhouette(). The input distance matrix is generated by cluster::daisy() using (complete cases of) horizon-level attributes from obj as named in vars. Note that observations with genhz labels specified in non.matching.code are removed filtered before calculation of the distance matrix.

Author