EcotoneFinder (version 0.2.3)

NetworkEco: Networks for ecotones and communities

Description

Networks for ecotones and communities

Usage

NetworkEco(ecotonefinder, threshold = 0.8, plot.type = c("percentage",
  "corrplot", "heatmap", "network"), method = c("cmeans", "vegclust"),
  dist.method = "inner_product", plot = c("species", "community"),
  order.sp = NULL, dist = c("count", "relative", "raw"),
  no.plot = FALSE, network.group = NULL, ...)

Arguments

ecotonefinder

A list containing elements named in the same way than EcotoneFinder function outcomes. Must contain cmeans results or vegclust results.

threshold

If count = T, the membership grade threshold used to sort the species in the different clusters.

plot.type

Which graphical representation to be plotted. Among "percentage", "corrplot", "heatmap","network".

method

The membership computation method to be used. One of "cmeans" or "vegclust". Must be present in the ecotonefinder list.

dist.method

Distance method for the computation of a distance matrix, when dist = "raw" and dist = "relative".

plot

If plot = "species", the distances are computed between the species in the data. If plot = "community", the distances are computed between the cluster centroids.

order.sp

Vector providing the order in which to arrange the species. If NULL, the column order will be kept.

dist

The type of data on which distance calculations are made from. If dist = "raw", the distance matrix is computed from the membership matrix directly. if dist = "relative", the distance matrix is computed from the relative memberships grades of each species in the clusters (between 0 and 1). If dist = "count", the species are assigned to clusters according to the threshold and the distance matrix is computed from the number of common species between the different clusters. See details.

no.plot

Logical. Should the plot be displayed?. Set to TRUE to gain computation time with large community matrix.

network.group

Grouping parameter for the network. Can be user defined (see qgraph documentation for details) but must be a factor of the same lenght as the nodes of the graph.

...

Additional arguments to be passed to the plotting functions, see details.

Value

A list containing the percentage matrix, the distance matrix and the network object (depending of the arguments passed to the function)

Details

The NetworkEco function provides wqys to explore the relations between fuzzy clusters. Several options are implemented. If dist = "raw", it computes a distance matrix from the membership grade matrix directly. If dist = "relative", the membership grades are standardized so that the sum of the membership grades of a given species equals to 1 for every points along the gradient (which corresponds to a percentage ot membership in each cluster). If dist = "count", the standardized membership grades of the are used to assign species in the community to a unique cluster and the number of common species between pairs of clusters is counted. The assignement of species to clusters is done by listing all the species that score a membership grade higher than the specified threshold in a cluster. The resulting list of species are then compared to one another.

The function also allows the computation of distances between species rather than between clusters, when plot = "species". This can only be done from the memberships grades (raw or relative) and this argument will be disregarded if dist = "count".

Several methods of visualisation are implemented: "percentage", "corrplot", "heatmap" and "network". If "percentage", a barplot (using ggplot2) of the standardized memberships grades per fuzzy cluster is plotted. It always plot the standardized membership grades regardless of the chosen dist option, but if dist = "count" or dist = "raw" are chosen, the function still compute the corresponding distance matrices and return them ti the output list. For time efficiency, it is not recommended to plot it when the number of species in the community is large (>100). "corrplot" and "heatmap" produce correlation matrix and heat map. The "network" is based on the qgraph function of the qgraph package. The ... argument may be used to pass additional arguments to the plotting functions (for graphical purposes).

Examples

# NOT RUN {
 #### Artificial dataset:
 SyntheticTrial <- SyntheticData(SpeciesNum = 21, CommunityNum = 3,
                                 SpCo = NULL, Length = 500,
                                 Parameters = list(a=rep(60, 3),
                                                   b=c(0,250,500),
                                                   c=rep(0.015,3)),
                                 pal = c("#008585", "#FBF2C4", "#C7522B"))

 ## Analyses:
 SyntheticEcoFinder <- EcotoneFinder(data = SyntheticTrial[,-1],
                                     dist = SyntheticTrial$Distance,
                                     method = "all",
                                     groups = 3, standardize = "hellinger",
                                     diversity = "all")

 ## Percentage plot:
 SyntheticNetwork <- NetworkEco(SyntheticEcoFinder, threshold = .3, method = "cmeans",
                                plot.type = "percentage", dist = "count")

 ## Heatmap plot:
 SyntheticNetwork <- NetworkEco(SyntheticEcoFinder, plot.type = "heatmap",
                                method = "cmeans", dist = "raw", plot = "species")

 ## Network:
 # From raw membership grades:
 SyntheticNetwork <- NetworkEco(SyntheticEcoFinder, plot.type = "network",
                                method = "cmeans", dist = "raw", plot = "species")

 # From number of species per clusters:
 SyntheticNetwork <- NetworkEco(SyntheticEcoFinder, plot.type = "network", threshold = .3,
                                method = "cmeans", dist = "count", plot = "community",
                                layout = "spring")

# }