# weighting-functions: Weighting Functions for Spatial Eigenvector Map In guenardg/codep: Multiscale Codependence Analysis

 weighting-functions R Documentation

## Weighting Functions for Spatial Eigenvector Map

### Description

A set of common distance weighting functions to calculate spatial eignevector maps using function `eigenmap`.

### Usage

``````wf.sqrd(d)

wf.RBF(d, wpar = 1)

wf.PCNM(d, boundaries, wpar = 4)

wf.binary(d, boundaries)

wf.Drayf1(d, boundaries)

wf.Drayf2(d, boundaries, wpar = 1)

wf.Drayf3(d, boundaries, wpar = 1)
``````

### Arguments

 `d` A triangular ('dist-class') or rectangular geographic distance matrix produced by `dist`, `Euclid`, or `geodesics`. `wpar` Where applicable, a parameter controlling the shape of the spatial weighting function. `boundaries` Where applicable, a two-element numeric vector containing the lower and upper threshold values used to obtain the connectivity matrix. (see details).

### Details

These functions are meant primarily to be called within functions `eigenmap` and `eigenmap.score`. In `eigenmap`, argument `d` is a lower-triangular 'dist-class' object and the resulting lower-triangular weight matrix is used in calculating the spatial eigenvector map. In `eigenmap.score`, `d` is a rectangular matrix of the distances between a set of arbitrary locations (rows) and reference locations (columns; the locations for which the the spatial eigenvector map has been built and the resulting rectangular weight matrix is used to calculate spatial eigenfunction values. These values allow one to use the spatial information of a data set for making predictions at arbitrary values.

'Wf.sqrd' (default value) consists in taking w_i,j = -0.5*d_i,j and does not involve any truncation.

'Wf.RBF' consists in taking w_i,j = exp(-wpar*d_i,j^2) and does not involve any truncation, where wpar is a non-zero real positive value (default: 1).

'Wf.binary' the spatial weighting matrix is simply the connectivity matrix.

'Wf.PCNM' is a_i,j = 1 - (d_i,j / (wpar*boundaries_2))^2, where wpar is a non-zero real positive value (default: 4).

'Wf.Drayf1' is a_i,j = 1 - (d_i,j / d_max) where d_max is the distance between the two most distant locations in the set.

'Wf.Drayf2' is a_i,j = 1 - (d_i,j / d_max)^wpar, where wpar is a non-zero real positive value (default: 1).

'Wf.Drayf3' is a_i,j = 1 / d_i,j^wpar, where wpar is a non-zero real positive value (default: 1).

Functions `Wf.Drayf1`, `Wf.Drayf2`, and `Wf.Drayf3` were proposed by Dray et al. (2006) and function `PCNM` was proposed by Legendre and Legendre (2012).

The `Wf.sqrd` weighting approach is equivalent to submitting the elementwise square-root of the distance matrix to a principal coordinate analysis. It was proposed by Diniz-Filho et al. (2013) and is equivalent, for evenly spaced transect or surfaces (square or rectangle), to using the basis functions of type II discrete cosine basis transforms; a fact that has gone unnoticed by Diniz-Filho et al. (2013).

The radial basis function (RBF) is a widespread kernel method involving sets of real-valued functions whose values depend on the distance between any given input coordinate and a set of fixed points (a single fixed point for each function). It is implemented using function `Wf.RBF` using all the sampling points as the fixed points.

When calculating the connectivity matrix, pairs of location whose distance to one another are between the boundary values (argument `bounraries`) are considered as neighbours (b_i,j=1) whereas values located below the minimum and above the maximum are considered as equivalent or distant, respectively (b_i,j=0 in both cases).

User may implement custom weighting functions. These functions must at the very least have an argument `d`, and can be given arguments `boundaries` and `wpar`. Argument `wpar` may be a vector with any number of elements. They should be added to the R-code file (weighting-functions.R). User-provided weighting functions with an argument `wpar` must come with a valid default value for that parameter since `eigenmap` may internally call it without a formal value.

### Value

A 'dist-class' object when argument `d` is a 'dist-class' object or a rectangular matrix when argument `d` is a rectangular matrix, either one with the weights as its values.

### Functions

• `wf.sqrd()`: Principal coordinates of the square-root distance matrix (Diniz-Filho et al. 2013).

• `wf.RBF()`: Radial basis functions with the observations as the kernels.

• `wf.PCNM()`: Borcard & Legendre's (2002) principal coordinates of the neighbour matrix approach.

• `wf.binary()`: Dray et al. (2006) Moran's eigenvector maps (distance-based binary connections without continuous weighting of the neighbours).

• `wf.Drayf1()`: Dray et al. (2006) Moran's eigenvector maps (distance-based binary connections with continuous weighting of the neighbours: f1).

• `wf.Drayf2()`: Dray et al. (2006) Moran's eigenvector maps (distance-based binary connections with continuous weighting of the neighbours: f2).

• `wf.Drayf3()`: Dray et al. (2006) Moran's eigenvector maps (distance-based binary connections with continuous weighting of the neighbours: f3).

### Author(s)

Guillaume Guenard and Pierre Legendre, Bertrand Pages Maintainer: Guillaume Guenard <guillaume.guenard@gmail.com>

### References

Borcard, D. and Legendre, P. 2002. All-scale spatial analysis of ecological data by means of principal coordinates of neighbour matrices. Ecol. Model. 153: 51-68

Diniz-Filho, J. A. F.; Diniz, J. V. B. P. L.; Rangel, T. F.; Soares, T. F.; de Campos Telles, M. P.; Garcia Collevatti, R. and Bini, L. M. 2013. A new eigenfunction spatial analysis describing population genetic structure. Genetica 141:479-489.

Dray, S.; Legendre, P. and Peres-Neto, P. 2006. Spatial modelling: a comprehensive framework for principal coordinate analysis of neighbor matrices (PCNM). Ecol. Modelling 196: 483-493

Legendre, P. and Legendre, L. 2012. Numerical Ecology, 3rd English edition. Elsevier Science B.V., Amsterdam, The Netherlands.

### Examples

``````locations <- c(1,2,4,7,10,14,17,21)
D <- dist(locations)
wf.sqrd(D)
wf.RBF(D, wpar = 0.1)
wf.binary(D, c(0,5))
wf.PCNM(D, c(0,5))
wf.Drayf1(D, c(0,5))
wf.Drayf2(D, c(0,5), 0.5)
wf.Drayf3(D, c(0,5), 0.5)

emap <- eigenmap(D, locations, wf.Drayf2, c(0,5), 0.5)
emap

emap <- eigenmap(D, locations, wf.Drayf3, c(0,5), 0.25)
emap

emap <- eigenmap(D, locations, wf.RBF, wpar = 0.1)
emap

``````

guenardg/codep documentation built on April 16, 2024, 9:01 p.m.