spatialsegregation-segregationFun: Spatial Segregation Function

Description

Compute the spatial exposure (segregation vs. mingling) features from a given multitype point pattern. Usage of shortcuts minglingF, isarF, shannonF and simpsonF highly recommended.

Usage

segregationFun(X, fun="isar", r=NULL, ntype="geometric", funpars=NULL, 
toroidal=FALSE, minusRange=TRUE,  included=NULL, dbg=FALSE, 
doDists=FALSE, prepRange=0.0, prepGraph=NULL, prepGraphIsTarget=FALSE,
weightMatrix=NULL, translate=FALSE)

Arguments

Multitype point pattern of class ppp (see package 'spatstat')

fun

Default "isar". Takes "isar","mingling","shannon" and "simpson", see below.

Vector for the neighbourhood defining graph, e.g. "geometric" graph with different ranges. See below.

ntype

Default "geometric". Type of the neighbourhood graph. Accepts: "knn", "geometric", "delauney", "gabriel".

funpars

Default NULL. Parameter(s) for the measure. Mingling: c(i,j), where i= only for type i (0 for all), j=1 -> ratio version. ISAR: i, i=type (integer). Shannon: 0 or 1, see v2 in shannonF. Simpson: none.

toroidal

Default FALSE. If TRUE, use a toroidal correction in distance calculation. Works at the moment only for rectangular windows and "geometric" or "knn" graph.

minusRange

If TRUE, adaptive minus-sampling is employed. Overrides included-vector. If given as a positive number, included-vector is created with points with distance atleast minusRange from the border.

included

boolean-vector of length |pp|. included[i]==TRUE => pp[i] included in calculations. Used for minus-sampling border correction.

dbg

Default FALSE. Print additional runtime texts.

doDists

Default TRUE. Precalculate distances for speed. Be aware of memory requirements, n*(n-1)!

prepRange

Default 0. If >0, shrink the search space for neighbourhoods by searching only points within distance R i.e. precalculates a geometric graph.

prepGraph

Precalculated graph for the point pattern. If given, The prepRange, dodists and toroidal are ignored and calculations are carried using the prepGraph as a starting point. Useful for huge datasets.

prepGraphIsTarget

If TRUE, precalculated graph prepGraph is used to calculate a single function value directly, all other neighbourhood parameters are ignored.

weightMatrix

See isarF for this.

translate

Use translation correction (see e.g. documentation of spatstat::Kest for details). Used only in mingling index.

Value

Returns an object of class fv, see spatstat for more details. Basically a list with the computed values and parameter values.

Details

This is the general function for computing the spatial exposure (segregation/mingling) features. Used by minglingF, shannonF, simpsonF and isarF, which should be preferred for better (and nicer) outcome.

Possible neighbourhood relations for the spatial version include geometric, k-nearest neighbours, Delauney, and Gabriel. Delauney and Gabriel are parameter free, so given r has no meaning. In geometric graph, r is a vector of distances (sizes of the surrounding 'disc') and for k-nn r is the vector of neighbourhood abundances for each point to consider in the calculation of the spatial exposure measures. The basic type of spatial summary uses range, or 'geometric' graph connections with varying neighbourhood parameter.

For geometric and knn, the calculations are done by shrinking the graph given by the largest value of r. If dealing with large datasets, it is advisable to give preprocessing range, prepRange. The algorithm first calculates a geometric graph with parameter prepRange, and uses this as basis for finding the needed neighbourhoods. Speeds up calculations. prepGraph, if given, works as the preprocessed geometric graph. But make sure prepRange is large enough (e.g. in geometric, prepRange>max(r)).

The doDists option speeds up calculations by precomputing the pairwise distances but takes n*(n-1) memory!

For border correction, use minusRange for reduced border correction (for rectangular windows only). If using geometric or knn neighbourhoods, the option toroidal for toroidal correction is also available. The vector included can be given for more specific minus-correction, only those points with TRUE (1) value are used in calculation. However, the neighbourhoods are calculated with all points.