mFD (version 1.0.7)

quality.fspaces: Compute functional spaces and their quality

Description

Compute a Principal Coordinates Analysis (PCoA) using functional distance between species. Then the function evaluates the quality of spaces built using an increasing number of principal components. Quality is evaluated as the (absolute or squared) deviation between trait-based distance (input) and distance in the PCoA-based space (raw Euclidean distance or scaled distance according to its maximum value and maximum of trait-based distance). Option to compute a functional dendrogram and its quality. This function is based on the framework presented in Maire et al. (2015).

Usage

quality.fspaces(
  sp_dist,
  fdendro = NULL,
  maxdim_pcoa = 10,
  deviation_weighting = "absolute",
  fdist_scaling = FALSE
)

Value

A list with:

• $quality_fspaces: a data frame with quality metric(s) for each functional space. Functional spaces are named as 'pcoa_.d' and if required 'tree_clustering method'. Quality metrics are named after deviation_weighting ('mad' for 'absolute' and and 'rmsd' for 'squared') and if fdist_scaling is TRUE with suffix '_scaled'.

• $details_trdist a list with 2 elements: $trdist_summary a vector with minimum (min), maximum (max), mean (mean) and standard deviation (sd) of sp_dist; $trdist_euclidean a logical value indicating whether sp_dist checks Euclidean properties.

• $details_fspaces a list with 4 elements: $sp_pc_coord a matrix with coordinates of species (rows) along Principal Components (columns) with positive eigenvalues ; $pc_eigenvalues a matrix with eigenvalues of axes from PCoA ; $dendro a hclust object with the dendrogram details (null if no dendrogram computed) ; $pairsp_fspaces_dist a dataframe containing for each pair of species (rows), their names in the 2 first columns ('sp.x' and 'sp.y'), their distance based on trait-values ('tr'), and their Euclidean (for PCoA) or cophenetic (for dendrogram if computed) distance in each of the functional space computed ('pcoa_1d', 'pcoa_2d', ... , 'tree_clust'); if fdist_scaling = TRUE, $pairsp_fspaces_dist_scaled a data frame with scaled values of distances in functional spaces.

• $details_deviation a list of data frames: $dev_distsp a dataframe containing for each space (columns) the difference for all species pairs (rows) of the distance in the functional space and trait-based distance (i.e. positive deviation indicates overestimation of actual distance) ; $abs_dev_distsp and/or $sqr_dev_distsp, dataframes with for each space (columns) and all species pairs (rows) the absolute or squared deviation of distance ; if fdist_scaling = TRUE $dev_distsp_scaled, and $abs_dev_distsp_scaled and/or $sqr_dev_distsp_scaled, data frames with deviation computed on scaled distance in functional spaces.

Arguments

sp_dist

a dist object with pairwise distance among all species (at least 3 species needed). Functional distance matrix from trait values can be computed using funct.dist function.

fdendro

a character string indicating the clustering algorithm to use to compute dendrogram. Should be one of the method recognized by hclust (e.g. 'average' for UPGMA). Default: fdendro = NULL (so no dendrogram computed).

maxdim_pcoa

a single numeric value with maximum number of PCoA axes to consider to build multidimensional functional spaces. Default: maxdim_pcoa = 10. See below about number of axes actually considered.

deviation_weighting

a character string referring to the method(s) used to weight the differences between species pairwise distance in the functional space and trait-based distance. 'absolute' (default) means absolute differences are used to compute mean absolute deviation mad index; 'squared' means squared differences are used to compute root of mean squared deviation rmsd index. Both values could be provided to compare quality metrics.

fdist_scaling

a vector with logical value(s) specifying whether distances in the functional space should be scaled before computing differences with trait-based distances. Scaling ensures that trait-based distances and distances in the functional space have the same maximum. Default: FALSE. Both values could be provided to compare quality metrics.

Author

Sebastien Villeger, Eva Maire, and Camille Magneville

References

Maire et al. (2015) How many dimensions are needed to accurately assess functional diversity? A pragmatic approach for assessing the quality of functional spaces Global Ecology and Biogeography, 24, 728-740.

Examples

# Load Species x Traits Data
data("fruits_traits", package = "mFD")

# Load Traits x Categories Data
data("fruits_traits_cat", package = "mFD")

# Compute Functional Distance
sp_dist_fruits <- mFD::funct.dist(
  sp_tr         = fruits_traits,
  tr_cat        = fruits_traits_cat,
  metric        = "gower",
  scale_euclid  = "scale_center",
  ordinal_var   = "classic",
  weight_type   = "equal",
  stop_if_NA    = TRUE)

# Compute Functional Spaces Quality (to retrieve species coordinates)
fspaces_quality_fruits <- mFD::quality.fspaces(
  sp_dist             = sp_dist_fruits,
  maxdim_pcoa         = 10,
  deviation_weighting = "absolute",
  fdist_scaling       = FALSE,
  fdendro             = "average")
 fspaces_quality_fruits
  
# Retrieve Species Coordinates
sp_faxes_coord_fruits <- fspaces_quality_fruits$details_fspaces$sp_pc_coord
sp_faxes_coord_fruits