0% found this document useful (0 votes)

55 views86 pages

GWmodel PDF

This package provides functions for performing geographically weighted regression and other local spatial statistical techniques. It includes functions for bandwidth selection, geographically weighted regression, principal components analysis, and discriminant analysis. The package is useful for situations where global models do not adequately describe spatial data and local models provide better fit. It allows for both basic and robust locally weighted versions of the implemented methods.

Uploaded by

Izzah Hilmy

We take content rights seriously. If you suspect this is your content, claim it here.

Available Formats

Download as PDF, TXT or read online on Scribd

0% found this document useful (0 votes)

55 views86 pages

GWmodel PDF

Uploaded by

Izzah Hilmy

We take content rights seriously. If you suspect this is your content, claim it here.

Available Formats

Download as PDF, TXT or read online on Scribd

You are on page 1/ 86

Package ‘GWmodel’

July 15, 2019

Type Package
Version 2.1-3
Date 2019-07-11
Title Geographically-Weighted Models
Depends R (>= 3.0.0),maptools (>= 0.5-2),
robustbase,sp,Rcpp,spatialreg
Imports methods, grDevices, stats,graphics,spacetime,spdep,FNN
LinkingTo Rcpp, RcppArmadillo
Suggests mvoutlier, RColorBrewer, gstat,spData
Description Techniques from a particular branch of spatial statistics,termed geographically-
weighted (GW) models. GW models suit situations when data are not de-
scribed well by some global model, but where there are spatial regions where a suitably lo-
calised calibration provides a better description. 'GWmodel' includes functions to cali-
brate: GW summary statistics (Brunsdon et al. 2002)<doi: 10.1016/s0198-9715(01)00009-
6>, GW principal components analysis (Har-
ris et al. 2011)<doi: 10.1080/13658816.2011.554838>, GW discriminant analysis (Bruns-
don et al. 2007)<doi: 10.1111/j.1538-4632.2007.00709.x> and various forms of GW regres-
sion (Brunsdon et al. 1996)<doi: 10.1111/j.1538-4632.1996.tb00936.x>; some of which are pro-
vided in basic and robust (outlier resistant) forms.
Author Binbin Lu[aut], Paul Harris[aut], Martin Charlton[aut], Chris Bruns-
don[aut], Tomoki Nakaya[aut], Daisuke Murakami[aut],Isabella Gollini[ctb]
Maintainer Binbin Lu <[email protected]>
License GPL (>= 2)
Repository CRAN

URL http://gwr.nuim.ie/
NeedsCompilation yes
Date/Publication 2019-07-15 08:10:13 UTC

1
2 R topics documented:

R topics documented:
GWmodel-package . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
bw.ggwr . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
bw.gtwr . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
bw.gwda . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
bw.gwpca . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
bw.gwr . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
bw.gwr.lcr . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
bw.gwss.average . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
DubVoter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
EWHP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
EWOutline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
Georgia . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
GeorgiaCounties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
ggwr.basic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
ggwr.cv . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
ggwr.cv.contrib . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
gtwr . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
gw.dist . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
gw.pcplot . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
gw.weight . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
gwda . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
gwpca . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
gwpca.check.components . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
gwpca.cv . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
gwpca.cv.contrib . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
gwpca.glyph.plot . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
gwpca.montecarlo.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
gwpca.montecarlo.2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
gwr.basic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
gwr.bootstrap . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
gwr.collin.diagno . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
gwr.cv . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
gwr.cv.contrib . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
gwr.hetero . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
gwr.lcr . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
gwr.lcr.cv . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
gwr.lcr.cv.contrib . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
gwr.mink.approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
gwr.mink.matrixview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
gwr.mink.pval . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
gwr.mixed . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
gwr.model.selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
gwr.model.sort . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
gwr.model.view . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
gwr.montecarlo . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
gwr.multiscale . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
GWmodel-package 3

gwr.predict . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
gwr.robust . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
gwr.scalable . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
gwr.t.adjust . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
gwr.write . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
gwss . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
gwss.montecarlo . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
LondonBorough . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
LondonHP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
USelect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82

Index 83

GWmodel-package Geographically-Weighted Models

Description
In GWmodel, we introduce techniques from a particular branch of spatial statistics, termed geographically-
weighted (GW) models. GW models suit situations when data are not described well by some global
model, but where there are spatial regions where a suitably localised calibration provides a better
description. GWmodel includes functions to calibrate: GW summary statistics, GW principal com-
ponents analysis, GW discriminant analysis and various forms of GW regression; some of which
are provided in basic and robust (outlier resistant) forms.

Details

Package: GWmodel
Type: Package
Version: 2.1-3
Date: 2019-07-11
License: GPL (>=2)
LazyLoad: yes

Note

Acknowledgements: We gratefully acknowledge support from Science Foundation Ireland under

the National Development Plan through the award of a Strategic Research Centre grant 07-SRC-
I1168.
Beta versions can always be found at https://github.com/lbb220/GWmodel, which includes all
the newly developed functions for GW models.
For latest tutorials on using GWmodel please go to: https://rpubs.com/gwmodel
4 bw.ggwr

Author(s)
Binbin Lu, Paul Harris, Martin Charlton, Chris Brunsdon, Tomoki Nakaya, Daisuke Murakami,Isabella
Gollini
Maintainer: Binbin Lu <[email protected]>

References
Gollini I, Lu B, Charlton M, Brunsdon C, Harris P (2015) GWmodel: an R Package for explor-
ing Spatial Heterogeneity using Geographically Weighted Models. Journal of Statistical Software,
63(17):1-50, http://www.jstatsoft.org/v63/i17/
Lu B, Harris P, Charlton M, Brunsdon C (2014) The GWmodel R Package: further topics for explor-
ing Spatial Heterogeneity using Geographically Weighted Models. Geo-spatial Information Science
17(2): 85-101, http://www.tandfonline.com/doi/abs/10.1080/10095020.2014.917453

bw.ggwr Bandwidth selection for generalised geographically weighted regres-

sion (GWR)

Description
A function for automatic bandwidth selection to calibrate a generalised GWR model

Usage
bw.ggwr(formula, data, family ="poisson", approach="CV",
kernel="bisquare",adaptive=FALSE, p=2, theta=0, longlat=F,dMat)

Arguments
formula Regression model formula of a formula object
data a Spatial*DataFrame, i.e. SpatialPointsDataFrame or SpatialPolygonsDataFrame
as defined in package sp
family a description of the error distribution and link function to be used in the model,
which can be specified by “poisson” or “binomial”
approach specified by CV for cross-validation approach or by AIC corrected (AICc) ap-
proach
kernel function chosen as follows:
gaussian: wgt = exp(-.5*(vdist/bw)^2);
exponential: wgt = exp(-vdist/bw);
bisquare: wgt = (1-(vdist/bw)^2)^2 if vdist < bw, wgt=0 otherwise;
tricube: wgt = (1-(vdist/bw)^3)^3 if vdist < bw, wgt=0 otherwise;
boxcar: wgt=1 if dist < bw, wgt=0 otherwise
adaptive if TRUE calculate an adaptive kernel where the bandwidth corresponds to the
number of nearest neighbours (i.e. adaptive distance); default is FALSE, where
a fixed kernel is found (bandwidth is a fixed distance)
bw.gtwr 5

p the power of the Minkowski distance, default is 2, i.e. the Euclidean distance
theta an angle in radians to rotate the coordinate system, default is 0
longlat if TRUE, great circle distances will be calculated
dMat a pre-specified distance matrix, it can be calculated by the function gw.dist

Value
Returns the adaptive or fixed distance bandwidth

Note
For a discontinuous kernel function, a bandwidth can be specified either as a fixed (constant) dis-
tance or as a fixed (constant) number of local data (i.e. an adaptive distance). For a continuous
kernel function, a bandwidth can be specified either as a fixed distance or as a ’fixed quantity that
reflects local sample size’ (i.e. still an ’adaptive’ distance but the actual local sample size will be the
sample size as functions are continuous). In practise a fixed bandwidth suits fairly regular sample
configurations whilst an adaptive bandwidth suits highly irregular sample configurations. Adaptive
bandwidths ensure sufficient (and constant) local information for each local calibration. This note
is applicable to all GW models

Author(s)
Binbin Lu <[email protected]>

bw.gtwr Bandwidth selection for GTWR

Description
A function for automatic bandwidth selection to calibrate a GTWR model

Usage
bw.gtwr(formula, data, obs.tv, approach="CV",kernel="bisquare",adaptive=FALSE,
p=2, theta=0, longlat=F,lamda=0.05,t.units = "auto",ksi=0, st.dMat,
verbose=T)

Arguments
formula Regression model formula of a formula object
data a Spatial*DataFrame, i.e. SpatialPointsDataFrame or SpatialPolygonsDataFrame
as defined in package sp
obs.tv a vector of time tags for each observation, which could be numeric or of POSIXlt
class
approach specified by CV for cross-validation approach or by AIC corrected (AICc) ap-
proach
6 bw.gtwr

kernel function chosen as follows:

gaussian: wgt = exp(-.5*(vdist/bw)^2);
exponential: wgt = exp(-vdist/bw);
bisquare: wgt = (1-(vdist/bw)^2)^2 if vdist < bw, wgt=0 otherwise;
tricube: wgt = (1-(vdist/bw)^3)^3 if vdist < bw, wgt=0 otherwise;
boxcar: wgt=1 if dist < bw, wgt=0 otherwise
adaptive if TRUE calculate an adaptive kernel where the bandwidth (bw) corresponds to
the number of nearest neighbours (i.e. adaptive distance); default is FALSE,
where a fixed kernel is found (bandwidth is a fixed distance)
p the power of the Minkowski distance, default is 2, i.e. the Euclidean distance
theta an angle in radians to rotate the coordinate system, default is 0
longlat if TRUE, great circle distances will be calculated
lamda an parameter between 0 and 1 for calculating spatio-temporal distance
t.units character string to define time unit
ksi an parameter between 0 and PI for calculating spatio-temporal distance, see de-
tails in Wu et al. (2014)
st.dMat a pre-specified spatio-temporal distance matrix
verbose logical variable to define whether show the selection procedure

Value
Returns the adaptive or fixed distance bandwidth

Note
The function is developed according to the articles by Huang et al. (2010) and Wu et al. (2014).

Author(s)
Binbin Lu <[email protected]>

References
Huang, B., Wu, B., & Barry, M. (2010). Geographically and temporally weighted regression for
modeling spatio-temporal variation in house prices. International Journal of Geographical Informa-
tion Science, 24, 383-401.
Wu, B., Li, R., & Huang, B. (2014). A geographically and temporally weighted autoregressive
model with application to housing prices. International Journal of Geographical Information Sci-
ence, 28, 1186-1204.
Fotheringham, A. S., Crespo, R., & Yao, J. (2015). Geographical and Temporal Weighted Regres-
sion (GTWR). Geographical Analysis, 47, 431-452.
bw.gwda 7

bw.gwda Bandwidth selection for GW Discriminant Analysis

Description
A function for automatic bandwidth selection for GW Discriminant Analysis using a cross-validation
approach only

Usage
bw.gwda(formula, data, COV.gw = T, prior.gw = T, mean.gw = T,
prior = NULL, wqda = F, kernel = "bisquare", adaptive
= FALSE, p = 2, theta = 0, longlat = F,dMat)

Arguments
formula Model formula of a formula object
data a Spatial*DataFrame for training, i.e. SpatialPointsDataFrame or SpatialPoly-
gonsDataFrame as defined in package sp
COV.gw if true, localised variance-covariance matrix is used for GW discriminant analy-
sis; otherwise, global variance-covariance matrix is used
mean.gw if true, localised mean is used for GW discriminant analysis; otherwise, global
mean is used
prior.gw if true, localised prior probability is used for GW discriminant analysis; other-
wise, fixed prior probability is used
prior a vector of given prior probability
wqda if TRUE, a weighted quadratic discriminant analysis will be applied; otherwise
a weighted linear discriminant analysis will be applied
kernel function chosen as follows:
gaussian: wgt = exp(-.5*(vdist/bw)^2);
exponential: wgt = exp(-vdist/bw);
bisquare: wgt = (1-(vdist/bw)^2)^2 if vdist < bw, wgt=0 otherwise;
tricube: wgt = (1-(vdist/bw)^3)^3 if vdist < bw, wgt=0 otherwise;
boxcar: wgt=1 if dist < bw, wgt=0 otherwise
adaptive if TRUE calculate an adaptive kernel where the bandwidth (bw) corresponds to
the number of nearest neighbours (i.e. adaptive distance); default is FALSE,
where a fixed kernel is found (bandwidth is a fixed distance)
p the power of the Minkowski distance, default is 2, i.e. the Euclidean distance
theta an angle in radians to rotate the coordinate system, default is 0
longlat if TRUE, great circle distances will be calculated
dMat a pre-specified distance matrix, it can be calculated by the function gw.dist
8 bw.gwpca

Value
Returns the adaptive or fixed distance bandwidth.

Author(s)
Binbin Lu <[email protected]>

bw.gwpca Bandwidth selection for Geographically Weighted Principal Compo-

nents Analysis (GWPCA)

Description
A function for automatic bandwidth selection to calibrate a basic or robust GWPCA via a cross-
validation approach only

Usage
bw.gwpca(data,vars,k=2, robust=FALSE,kernel="bisquare",adaptive=FALSE,p=2,
theta=0, longlat=F,dMat)

adaptive if TRUE calculate an adaptive kernel where the bandwidth corresponds to the
number of nearest neighbours (i.e. adaptive distance); default is FALSE, where
a fixed kernel is found (bandwidth is a fixed distance)
p the power of the Minkowski distance, default is 2, i.e. the Euclidean distance
theta an angle in radians to rotate the coordinate system, default is 0
longlat if TRUE, great circle distances will be calculated
dMat a pre-specified distance matrix, it can be calculated by the function gw.dist

Value
Returns the adaptive or fixed distance bandwidth

Author(s)
Binbin Lu <[email protected]>

References
Harris P, Clarke A, Juggins S, Brunsdon C, Charlton M (2015) Enhancements to a geographically
weighted principal components analysis in the context of an application to an environmental data
set. Geographical Analysis 47: 146-172

bw.gwr Bandwidth selection for basic GWR

Description
A function for automatic bandwidth selection to calibrate a basic GWR model

Usage
bw.gwr(formula, data, approach="CV",kernel="bisquare",
adaptive=FALSE, p=2, theta=0, longlat=F,dMat)
10 bw.gwr

Arguments

formula Regression model formula of a formula object

data a Spatial*DataFrame, i.e. SpatialPointsDataFrame or SpatialPolygonsDataFrame
as defined in package sp
approach specified by CV for cross-validation approach or by AIC corrected (AICc) ap-
proach
kernel function chosen as follows:
gaussian: wgt = exp(-.5*(vdist/bw)^2);
exponential: wgt = exp(-vdist/bw);
bisquare: wgt = (1-(vdist/bw)^2)^2 if vdist < bw, wgt=0 otherwise;
tricube: wgt = (1-(vdist/bw)^3)^3 if vdist < bw, wgt=0 otherwise;
boxcar: wgt=1 if dist < bw, wgt=0 otherwise
adaptive if TRUE calculate an adaptive kernel where the bandwidth (bw) corresponds to
the number of nearest neighbours (i.e. adaptive distance); default is FALSE,
where a fixed kernel is found (bandwidth is a fixed distance)
p the power of the Minkowski distance, default is 2, i.e. the Euclidean distance
theta an angle in radians to rotate the coordinate system, default is 0
longlat if TRUE, great circle distances will be calculated
dMat a pre-specified distance matrix, it can be calculated by the function gw.dist

Value

Returns the adaptive or fixed distance bandwidth

Note

For a discontinuous kernel function, a bandwidth can be specified either as a fixed (constant) dis-
tance or as a fixed (constant) number of local data (i.e. an adaptive distance). For a continuous
kernel function, a bandwidth can be specified either as a fixed distance or as a ’fixed quantity that
reflects local sample size’ (i.e. still an ’adaptive’ distance but the actual local sample size will be the
sample size as functions are continuous). In practise a fixed bandwidth suits fairly regular sample
configurations whilst an adaptive bandwidth suits highly irregular sample configurations. Adaptive
bandwidths ensure sufficient (and constant) local information for each local calibration. This note
is applicable to all GW models

Author(s)

Binbin Lu <[email protected]>
bw.gwr.lcr 11

bw.gwr.lcr Bandwidth selection for locally compensated ridge GWR (GWR-LCR)

Description
A function for automatic bandwidth selection for gwr.lcr via a cross-validation approach only

Usage
bw.gwr.lcr(formula, data, kernel="bisquare",
lambda=0,lambda.adjust=FALSE,cn.thresh=NA,
adaptive=FALSE, p=2, theta=0, longlat=F,dMat)

Value
Returns the adaptive or fixed distance bandwidth
12 bw.gwss.average

Author(s)
Binbin Lu <[email protected]>

bw.gwss.average Bandwidth selection for GW summary averages

Description
A function for automatic bandwidth selections to calculate GW summary averages, including means
and medians, via a cross-validation approach.

Usage
bw.gwss.average(data, summary.locat, vars, kernel = "bisquare", adaptive = FALSE,
p = 2, theta = 0, longlat = F, dMat)

adaptive if TRUE calculate an adaptive kernel where the bandwidth (bw) corresponds to
the number of nearest neighbours (i.e. adaptive distance); default is FALSE,
where a fixed kernel is found (bandwidth is a fixed distance)
p the power of the Minkowski distance, default is 2, i.e. the Euclidean distance
theta an angle in radians to rotate the coordinate system, default is 0
longlat if TRUE, great circle distances will be calculated
dMat a pre-specified distance matrix, it can be calculated by the function gw.dist

Value
Returns the adaptive or fixed distance bandwidths (in a two-column matrix) for calculating the
averages of each variable.

Author(s)
Binbin Lu <[email protected]>

DubVoter Voter turnout data in Greater Dublin(SpatialPolygonsDataFrame)

Description
Voter turnout and social characters data in Greater Dublin for the 2002 General election and the
2002 census. Note that this data set was originally thought to relate to 2004, so for continuity we
have retained the associated variable names.

Usage
data(DubVoter)

Format
A SpatialPolygonsDataFrame with 322 electoral divisions on the following 11 variables.
DED_ID a vector of ID
X a numeric vector of x coordinates
Y a numeric vector of y coordinates
DiffAdd percentage of the population in each ED who are one-year migrants (i.e. moved to a
different address 1 year ago)
LARent percentage of the population in each ED who are local authority renters
SC1 percentage of the population in each ED who are social class one (high social class)
Unempl percentage of the population in each ED who are unemployed
LowEduc percentage of the population in each ED who are with little formal education
Age18_24 percentage of the population in each ED who are age group 18-24
Age25_44 percentage of the population in each ED who are age group 25-44
Age45_64 percentage of the population in each ED who are age group 45-64
GenEl2004 percentage of population in each ED who voted in 2004 election
14 EWHP

Details
Variables are from DubVoter.shp.

References
Kavanagh A (2006) Turnout or turned off? Electoral participation in Dublin in the early 21st Cen-
tury. Journal of Irish Urban Studies 3(2):1-24
Harris P, Brunsdon C, Charlton M (2011) Geographically weighted principal components analysis.
International Journal of Geographical Information Science 25 (10):1717-1736

Examples

data(DubVoter)
ls()
## Not run:
spplot(Dub.voter,names(Dub.voter)[4:12])

## End(Not run)

EWHP House price data set (DataFrame) in England and Wales

Description
A house price data set for England and Wales from 2001 with 9 hedonic (explanatory) variables.

Usage
data(EWHP)

Format
A data frame with 519 observations on the following 12 variables.
Easting a numeric vector, X coordinate
Northing a numeric vector, Y coordinate
PurPrice a numeric vector, the purchase price of the property
BldIntWr a numeric vector, 1 if the property was built during the world war, 0 otherwise
BldPostW a numeric vector, 1 if the property was built after the world war, 0 otherwise
Bld60s a numeric vector, 1 if the property was built between 1960 and 1969, 0 otherwise
Bld70s a numeric vector, 1 if the property was built between 1970 and 1979, 0 otherwise
Bld80s a numeric vector, 1 if the property was built between 1980 and 1989, 0 otherwise
TypDetch a numeric vector, 1 if the property is detached (i.e. it is a stand-alone house), 0 otherwise
TypSemiD a numeric vector, 1 if the property is semi detached, 0 otherwise
TypFlat a numeric vector, if the property is a flat (or ’apartment’ in the USA), 0 otherwise
FlrArea a numeric vector, floor area of the property in square metres
EWOutline 15

Author(s)
Binbin Lu <[email protected]>

References
Fotheringham, A.S., Brunsdon, C., and Charlton, M.E. (2002), Geographically Weighted Regres-
sion: The Analysis of Spatially Varying Relationships, Chichester: Wiley.

Examples
###
data(EWHP)
head(ewhp)
houses.spdf <- SpatialPointsDataFrame(ewhp[, 1:2], ewhp)
####Get the border of England and Wales
data(EWOutline)
plot(ewoutline)
plot(houses.spdf, add = TRUE, pch = 16)

EWOutline Outline of England and Wales for data EWHP

Description
Outline (SpatialPolygonsDataFrame) of the England and Wales house price data EWHP.

Usage
data(EWOutline)

Author(s)
Binbin Lu <[email protected]>

Georgia Georgia census data set (csv file)

Description
Census data from the county of Georgia, USA

Usage
data(Georgia)
16 GeorgiaCounties

Format
A data frame with 159 observations on the following 13 variables.
AreaKey An identification number for each county
Latitude The latitude of the county centroid
Longitud The longitude of the county centroid
TotPop90 Population of the county in 1990
PctRural Percentage of the county population defined as rural
PctBach Percentage of the county population with a bachelors degree
PctEld Percentage of the county population aged 65 or over
PctFB Percentage of the county population born outside the US
PctPov Percentage of the county population living below the poverty line
PctBlack Percentage of the county population who are black
ID a numeric vector of IDs
X a numeric vector of x coordinates
Y a numeric vector of y coordinates

Details
This data set can also be found in GWR 3 and in spgwr.

References
Fotheringham S, Brunsdon, C, and Charlton, M (2002), Geographically Weighted Regression: The
Analysis of Spatially Varying Relationships, Chichester: Wiley.

Examples
data(Georgia)
ls()
coords <- cbind(Gedu.df$X, Gedu.df$Y)
educ.spdf <- SpatialPointsDataFrame(coords, Gedu.df)
spplot(educ.spdf, names(educ.spdf)[4:10])

GeorgiaCounties Georgia counties data (SpatialPolygonsDataFrame)

Description
The Georgia census data with boundaries for mapping

Usage
data(GeorgiaCounties)
ggwr.basic 17

Details

This data set can also be found in GWR 3 and in spgwr.

Examples
data(GeorgiaCounties)
plot(Gedu.counties)
data(Georgia)
coords <- cbind(Gedu.df$X, Gedu.df$Y)
educ.spdf <- SpatialPointsDataFrame(coords, Gedu.df)
plot(educ.spdf, add=TRUE)

ggwr.basic Generalised GWR models with Poisson and Binomial options

Description

This function implements generalised GWR

Usage

ggwr.basic(formula, data, regression.points, bw, family =

"poisson", kernel = "bisquare", adaptive = FALSE, cv =
T, tol = 1e-05, maxiter = 20, p = 2, theta = 0,
longlat = F, dMat, dMat1)

## S3 method for class 'ggwrm'

print(x, ...)

Arguments

formula Regression model formula of a formula object

data a Spatial*DataFrame, i.e. SpatialPointsDataFrame or SpatialPolygonsDataFrame
as defined in package sp
regression.points
a Spatial*DataFrame object, i.e. SpatialPointsDataFrame or SpatialPolygons-
DataFrame as defined in package sp
bw bandwidth used in the weighting function, possibly calculated by bw.ggwr();fixed
(distance) or adaptive bandwidth(number of nearest neighbours)
family a description of the error distribution and link function to be used in the model,
which can be specified by “poisson” or “binomial”
18 ggwr.basic

kernel function chosen as follows:

gaussian: wgt = exp(-.5*(vdist/bw)^2);
exponential: wgt = exp(-vdist/bw);
bisquare: wgt = (1-(vdist/bw)^2)^2 if vdist < bw, wgt=0 otherwise;
tricube: wgt = (1-(vdist/bw)^3)^3 if vdist < bw, wgt=0 otherwise;
boxcar: wgt=1 if dist < bw, wgt=0 otherwise
adaptive if TRUE calculate an adaptive kernel where the bandwidth corresponds to the
number of nearest neighbours (i.e. adaptive distance); default is FALSE, where
a fixed kernel is found (bandwidth is a fixed distance)
cv if TRUE, cross-validation data will be calculated
tol the threshold that determines the convergence of the IRLS procedure
maxiter the maximum number of times to try the IRLS procedure
p the power of the Minkowski distance, default is 2, i.e. the Euclidean distance
theta an angle in radians to rotate the coordinate system, default is 0
longlat if TRUE, great circle distances will be calculated
dMat a pre-specified distance matrix between regression points and observations, it
can be calculated by the function gw.dist
dMat1 a square distance matrix between each pair of observations, it can be calculated
by the function gw.dist
x an object of class “ggwrm”, returned by the function gwr.generalised
... arguments passed through (unused)

Value
A list of class “ggwrm”:
GW.arguments a list class object including the model fitting parameters for generating the report
file
GW.diagnostic a list class object including the diagnostic information of the model fitting
glm.res an object of class inheriting from “glm” which inherits from the class “lm”, see
glm.
SDF a SpatialPointsDataFrame (may be gridded) or SpatialPolygonsDataFrame ob-
ject (see package “sp”) integrated with fit.points,GWR coefficient estimates, y
value,predicted values, coefficient standard errors and t-values in its "data" slot.
CV a data vector consisting of the cross-validation data

Note
Note that this function calibrates a Generalised GWR model via an approximating algorithm, which
is different from the back-fitting algorithm used in the GWR4 software by Tomoki Nakaya.

Author(s)
Binbin Lu <[email protected]>
ggwr.cv 19

References
Nakaya, T., A. S. Fotheringham, C. Brunsdon & M. Charlton (2005) Geographically weighted
Poisson regression for disease association mapping. Statistics in Medicine, 24, 2695-2717.
Nakaya, T., M. Charlton, S. Fotheringham & C. Brunsdon. 2009. How to use SGWRWIN (GWR4.0).
Maynooth, Ireland: National Centre for Geocomputation.
Fotheringham S, Brunsdon, C, and Charlton, M (2002), Geographically Weighted Regression: The
Analysis of Spatially Varying Relationships, Chichester: Wiley.

Examples
data(LondonHP)
## Not run:
DM<-gw.dist(dp.locat=coordinates(londonhp))
bw.f1 <- bw.ggwr(BATH2~FLOORSZ,data=londonhp, dMat=DM)
res.poisson<-ggwr.basic(BATH2~FLOORSZ, bw=bw.f1,data=londonhp, dMat=DM)
bw.f2 <- bw.ggwr(BATH2~FLOORSZ,data=londonhp, dMat=DM,family ="binomial")
res.binomial<-ggwr.basic(BATH2~FLOORSZ, bw=bw.f2,data=londonhp, dMat=DM,
family ="binomial")

## End(Not run)

ggwr.cv Cross-validation score for a specified bandwidth for generalised GWR

Description
This function finds the cross-validation score for a specified bandwidth for generalised GWR. It can
be used to construct the bandwidth function across all possible bandwidths and compared to that
found automatically.

Usage
ggwr.cv(bw, X, Y,family="poisson", kernel="bisquare",adaptive=F, dp.locat,
p=2, theta=0, longlat=F,dMat)

kernel function chosen as follows:

Value
CV.score cross-validation score

Author(s)
Binbin Lu <[email protected]>

ggwr.cv.contrib Cross-validation data at each observation location for a generalised

GWR model

Description
This function finds the individual cross-validation score at each observation location, for a gener-
alised GWR model, for a specified bandwidth. These data can be mapped to detect unusually high
or low cross-validations scores.

Usage
ggwr.cv.contrib(bw, X, Y,family="poisson", kernel="bisquare",adaptive=F,
dp.locat, p=2, theta=0, longlat=F,dMat)

family a description of the error distribution and link function to be used in the model,
which can be specified by “poisson” or “binomial”
kernel function chosen as follows:
gaussian: wgt = exp(-.5*(vdist/bw)^2);
exponential: wgt = exp(-vdist/bw);
bisquare: wgt = (1-(vdist/bw)^2)^2 if vdist < bw, wgt=0 otherwise;
tricube: wgt = (1-(vdist/bw)^3)^3 if vdist < bw, wgt=0 otherwise;
boxcar: wgt=1 if dist < bw, wgt=0 otherwise
adaptive if TRUE calculate an adaptive kernel where the bandwidth (bw) corresponds to
the number of nearest neighbours (i.e. adaptive distance); default is FALSE,
where a fixed kernel is found (bandwidth is a fixed distance)
dp.locat a two-column numeric array of observation coordinates
p the power of the Minkowski distance, default is 2, i.e. the Euclidean distance
theta an angle in radians to rotate the coordinate system, default is 0
longlat if TRUE, great circle distances will be calculated
dMat a pre-specified distance matrix, it can be calculated by the function gw.dist

Value

CV a data vector consisting of squared residuals, whose sum is the cross-validation

score for the specified bandwidth

Author(s)

Binbin Lu <[email protected]>

gtwr Geographically and Temporally Weighted Regression

Description

A function for calibrating a Geographically and Temporally Weighted Regression (GTWR) model.

Usage

gtwr(formula, data, regression.points, obs.tv, reg.tv, st.bw, kernel="bisquare",

adaptive=FALSE, p=2, theta=0, longlat=F,lamda=0.05,t.units = "auto",ksi=0,
st.dMat)
22 gtwr

Arguments
formula Regression model formula of a formula object
data a Spatial*DataFrame, i.e. SpatialPointsDataFrame or SpatialPolygonsDataFrame
as defined in package sp
regression.points
a Spatial*DataFrame object, i.e. SpatialPointsDataFrame or SpatialPolygons-
DataFrame as defined in package sp; Note that no diagnostic information will
returned if it is assigned
obs.tv a vector of time tags for each observation, which could be numeric or of POSIXlt
class
reg.tv a vector of time tags for each regression location, which could be numeric or of
POSIXlt class
st.bw spatio-temporal bandwidth used in the weighting function, possibly calculated
by bw.gwr;fixed (distance) or adaptive bandwidth(number of nearest neighbours)
kernel function chosen as follows:
gaussian: wgt = exp(-.5*(vdist/bw)^2);
exponential: wgt = exp(-vdist/bw);
bisquare: wgt = (1-(vdist/bw)^2)^2 if vdist < bw, wgt=0 otherwise;
tricube: wgt = (1-(vdist/bw)^3)^3 if vdist < bw, wgt=0 otherwise;
boxcar: wgt=1 if dist < bw, wgt=0 otherwise
adaptive if TRUE calculate an adaptive kernel where the bandwidth (bw) corresponds to
the number of nearest neighbours (i.e. adaptive distance); default is FALSE,
where a fixed kernel is found (bandwidth is a fixed distance)
p the power of the Minkowski distance, default is 2, i.e. the Euclidean distance
theta an angle in radians to rotate the coordinate system, default is 0
longlat if TRUE, great circle distances will be calculated
lamda an parameter between 0 and 1 for calculating spatio-temporal distance
t.units character string to define time unit
ksi an parameter between 0 and PI for calculating spatio-temporal distance, see de-
tails in Wu et al. (2014)
st.dMat a pre-specified spatio-temporal distance matrix

Value
A list of class “gtwrm”:
GTW.arguments a list class object including the model fitting parameters for generating the report
file
GTW.diagnostic a list class object including the diagnostic information of the model fitting
lm an object of class inheriting from “lm”, see lm.
SDF a SpatialPointsDataFrame (may be gridded) or SpatialPolygonsDataFrame ob-
ject (see package “sp”) integrated with fit.points, GTWR coefficient estimates, y
value,predicted values, coefficient standard errors and t-values in its "data" slot.
timings starting and ending time.
this.call the function call used.
gw.dist 23

Note
The function implements GTWR model proposed by Huang et al. (2010) and Wu et al. (2014).

Author(s)
Binbin Lu <[email protected]>

gw.dist Distance matrix calculation

Description
Calculate a distance vector(matrix) between any GW model calibration point(s) and the data points.

Usage
gw.dist(dp.locat, rp.locat, focus=0, p=2, theta=0, longlat=F)

Arguments
dp.locat a numeric matrix of two columns giving the coordinates of the data points
rp.locat a numeric matrix of two columns giving the coordinates of the GW model cali-
bration points
focus an integer, indexing to the current GW model point, if focus=0, all the distances
between all the GW model calibration points and data points will be calculated
and a distance matrix will be returned; if 0<focus<length(rp.locat), then the dis-
tances between the ’focus’th GW model points and data points will be calculated
and a distance vector will be returned
p the power of the Minkowski distance, default is 2, i.e. the Euclidean distance
theta an angle in radians to rotate the coordinate system, default is 0
longlat if TRUE, great circle distances will be calculated
24 gw.pcplot

Value
Returns a numeric distance matrix or vector; matrix with its rows corresponding to the observations
and its columns corresponds to the GW model calibration points.

Author(s)
Binbin Lu <[email protected]>

gw.pcplot Geographically weighted parallel coordinate plot for investigating

multivariate data sets

Description
This function provides a geographically weighted parallel coordinate plot for locally investigating
a multivariate data set. It has an option that weights the lines of the plot with increasing levels of
transparency, according to their observation’s distance from a specified focal/observation point.
gw.pcplot 25

Usage

gw.pcplot(data,vars,focus,bw,adaptive = FALSE, ylim=NULL,ylab="",fixtrans=FALSE,

p=2, theta=0, longlat=F,dMat,...)

Arguments

data a Spatial*DataFrame, i.e. SpatialPointsDataFrame or SpatialPolygonsDataFrame

as defined in package sp
vars a vector of variable names to be evaluated
focus an integer, indexing to the observation point
bw bandwidth used in the weighting function;fixed (distance) or adaptive band-
width(number of nearest neighbours)
adaptive if TRUE calculate an adaptive kernel where the bandwidth (bw) corresponds to
the number of nearest neighbours (i.e. adaptive distance); default is FALSE,
where a fixed kernel is found (bandwidth is a fixed distance)
ylim the y limits of the plot
ylab a label for the y axis
fixtrans if TRUE, the transparency of the neighbouring observation plot lines increases
with distance; If FALSE a standard (non-spatial) parallel coordinate plot is re-
turned.
p the power of the Minkowski distance, default is 2, i.e. the Euclidean distance
theta an angle in radians to rotate the coordinate system, default is 0
longlat if TRUE, great circle distances will be calculated
dMat a pre-specified distance matrix, it can be calculated by the function gw.dist
... other graphical parameters, (see par)

Author(s)

Binbin Lu <[email protected]>

References

Harris P, Brunsdon C, Charlton M, Juggins S, Clarke A (2014) Multivariate spatial outlier detection
using robust geographically weighted methods. Mathematical Geosciences 46(1) 1-31
Harris P, Clarke A, Juggins S, Brunsdon C, Charlton M (2015) Enhancements to a geographically
weighted principal components analysis in the context of an application to an environmental data
set. Geographical Analysis 47: 146-172
26 gw.weight

gw.weight Weight matrix calculation

Description
Calculate a weight vector(matrix) from a distance vector(matrix).

Usage
gw.weight(vdist,bw,kernel,adaptive=FALSE)

Arguments
vdist a distance matrix or vector
bw bandwidth used in the weighting function, possibly calculated by bw.gwr;fixed
(distance) or adaptive bandwidth(number of nearest neighbours)
kernel function chosen as follows:
gaussian: wgt = exp(-.5*(vdist/bw)^2);
exponential: wgt = exp(-vdist/bw);
bisquare: wgt = (1-(vdist/bw)^2)^2 if vdist < bw, wgt=0 otherwise;
tricube: wgt = (1-(vdist/bw)^3)^3 if vdist < bw, wgt=0 otherwise;
boxcar: wgt=1 if dist < bw, wgt=0 otherwise
adaptive if TRUE calculate an adaptive kernel where the bandwidth (bw) corresponds to
the number of nearest neighbours (i.e. adaptive distance); default is FALSE,
where a fixed kernel is found (bandwidth is a fixed distance)

Value
Returns a numeric weight matrix or vector; matrix with its rows corresponding to the observations
and its columns corresponds to the GW model calibration points.

Note
The gaussian and exponential kernel functions are continuous and valued in the interval (0,1]; while
bisquare, tricube and boxcar kernel functions are discontinuous and valued in the interval [0,1].
Notably, the upper limit of the bandwidth is exactly the number of observations when the adaptive
kernel is used. In this function, the adaptive bandwidth will be specified as the number of observa-
tions even though a larger number is assigned. The function will be the same as a global application
function (i.e. all weights are 1) when the adaptive bandwidth is equal to or larger than the number
of observations when using the boxcar kernel function.

Author(s)
Binbin Lu <[email protected]>
gwda 27

gwda GW Discriminant Analysis

Description
This function implements GW discriminant analysis.

Usage
gwda(formula, data, predict.data,validation = T, COV.gw=T,
mean.gw=T, prior.gw=T, prior=NULL, wqda =F,
kernel = "bisquare", adaptive = FALSE, bw,
p = 2, theta = 0, longlat = F,dMat)
## S3 method for class 'gwda'
print(x, ...)

Arguments
formula Model formula of a formula object
data a Spatial*DataFrame for training, i.e. SpatialPointsDataFrame or SpatialPoly-
gonsDataFrame as defined in package sp
predict.data a Spatial*DataFrame object for prediction, i.e. SpatialPointsDataFrame or Spa-
tialPolygonsDataFrame as defined in package sp; if it is not given, the traing
data will be predicted using leave-one-out cross-validation.
validation If TRUE, the results from the prediction will be validated and the correct pro-
portion will be calculated.
COV.gw if true, localised variance-covariance matrix is used for GW discriminant analy-
sis; otherwise, global variance-covariance matrix is used
mean.gw if true, localised mean is used for GW discriminant analysis; otherwise, global
mean is used
prior.gw if true, localised prior probability is used for GW discriminant analysis; other-
wise, fixed prior probability is used
prior a vector of given prior probability
wqda if TRUE, weighted quadratic discriminant analysis will be applied; otherwise
weighted linear discriminant analysis will be applied
kernel function chosen as follows:
gaussian: wgt = exp(-.5*(vdist/bw)^2);
exponential: wgt = exp(-vdist/bw);
bisquare: wgt = (1-(vdist/bw)^2)^2 if vdist < bw, wgt=0 otherwise;
tricube: wgt = (1-(vdist/bw)^3)^3 if vdist < bw, wgt=0 otherwise;
boxcar: wgt=1 if dist < bw, wgt=0 otherwise
adaptive if TRUE calculate an adaptive kernel where the bandwidth corresponds to the
number of nearest neighbours (i.e. adaptive distance); default is FALSE, where
a fixed kernel is found (bandwidth is a fixed distance)
28 gwpca

bw bandwidth used in the weighting function, possibly calculated by bw.gwpca;fixed

(distance) or adaptive bandwidth(number of nearest neighbours)
p the power of the Minkowski distance, default is 2, i.e. the Euclidean distance
theta an angle in radians to rotate the coordinate system, default is 0
longlat if TRUE, great circle distances will be calculated
dMat a pre-specified distance matrix, it can be calculated by the function gw.dist
x an object of class “gwda”
... arguments passed through (unused)

Value

A class of object “gwda”

Author(s)

Binbin Lu <[email protected]>

References

Brunsdon, C, Fotheringham S, and Charlton, M (2007), Geographically Weighted Discriminant

Analysis, Geographical Analysis 39:376-396
Lu B, Harris P, Charlton M, Brunsdon C (2014) The GWmodel R Package: further topics for
exploring Spatial Heterogeneity using Geographically Weighted Models. Geo-spatial Information
Science 17(2): 85-101

gwpca GWPCA

Description

This function implements basic or robust GWPCA.

Usage

gwpca(data, elocat, vars, k = 2, robust = FALSE, kernel = "bisquare",

adaptive = FALSE, bw, p = 2, theta = 0, longlat = F, cv = T, scores=F,
dMat)
## S3 method for class 'gwpca'
print(x, ...)
gwpca 29

Arguments
data a Spatial*DataFrame, i.e. SpatialPointsDataFrame or SpatialPolygonsDataFrame
as defined in package sp
elocat a two-column numeric array or Spatial*DataFrame object for providing evalu-
ation locations, i.e. SpatialPointsDataFrame or SpatialPolygonsDataFrame as
defined in package sp
vars a vector of variable names to be evaluated
k the number of retained components; k must be less than the number of variables
robust if TRUE, robust GWPCA will be applied; otherwise basic GWPCA will be ap-
plied
kernel function chosen as follows:
gaussian: wgt = exp(-.5*(vdist/bw)^2);
exponential: wgt = exp(-vdist/bw);
bisquare: wgt = (1-(vdist/bw)^2)^2 if vdist < bw, wgt=0 otherwise;
tricube: wgt = (1-(vdist/bw)^3)^3 if vdist < bw, wgt=0 otherwise;
boxcar: wgt=1 if dist < bw, wgt=0 otherwise
adaptive if TRUE calculate an adaptive kernel where the bandwidth corresponds to the
number of nearest neighbours (i.e. adaptive distance); default is FALSE, where
a fixed kernel is found (bandwidth is a fixed distance)
bw bandwidth used in the weighting function, possibly calculated by bw.gwpca;fixed
(distance) or adaptive bandwidth(number of nearest neighbours)
p the power of the Minkowski distance, default is 2, i.e. the Euclidean distance
theta an angle in radians to rotate the coordinate system, default is 0
longlat if TRUE, great circle distances will be calculated
cv If TRUE, cross-validation data will be found that are used to calculate the cross-
validation score for the specified bandwidth.
scores if scores = TRUE, the scores of the supplied data on the principal components
will be calculated.
dMat a pre-specified distance matrix, it can be calculated by the function gw.dist
x an object of class “gwpca”, returned by the function gwpca
... arguments passed through (unused)

Value
A list of class “gwpca”:

GW.arguments a list class object including the model fitting parameters for generating the report
file
pca an object of class inheriting from “princomp”, see princomp.
loadings the localised loadings
30 gwpca

SDF a SpatialPointsDataFrame (may be gridded) or SpatialPolygonsDataFrame ob-

ject (see package “sp”) integrated with local proportions of variance for each
principle components, cumulative proportion and winning variable for the 1st
principle component in its "data" slot.
gwpca.scores the localised scores of the supplied data on the principal components
var The local amount of variance accounted for by each component
CV Vector of cross-validation data
timings starting and ending time.

Author(s)
Binbin Lu <[email protected]>

References
Fotheringham S, Brunsdon, C, and Charlton, M (2002), Geographically Weighted Regression: The
Analysis of Spatially Varying Relationships, Chichester: Wiley.
Harris P, Brunsdon C, Charlton M (2011) Geographically weighted principal components analysis.
International Journal of Geographical Information Science 25:1717-1736
Harris P, Brunsdon C, Charlton M, Juggins S, Clarke A (2014) Multivariate spatial outlier detection
using robust geographically weighted methods. Mathematical Geosciences 46(1) 1-31
Harris P, Clarke A, Juggins S, Brunsdon C, Charlton M (2014) Geographically weighted meth-
ods and their use in network re-designs for environmental monitoring. Stochastic Environmental
Research and Risk Assessment 28: 1869-1887
Harris P, Clarke A, Juggins S, Brunsdon C, Charlton M (2015) Enhancements to a geographically
weighted principal components analysis in the context of an application to an environmental data
set. Geographical Analysis 47: 146-172

Examples
## Not run:
if(require("mvoutlier") && require("RColorBrewer"))
{
data(bsstop)
Data.1 <- bsstop[, 1:14]
colnames(Data.1)
Data.1.scaled <- scale(as.matrix(Data.1[5:14])) # standardised data...
rownames(Data.1.scaled) <- Data.1[, 1]
#compute principal components:
pca <- princomp(Data.1.scaled, cor = FALSE, scores = TRUE)
# use covariance matrix to match the following...
pca$loadings
data(bss.background)
backdrop <- function()
plot(bss.background, asp = 1, type = "l", xaxt = "n", yaxt = "n",
xlab = "", ylab = "", bty = "n", col = "grey")
pc1 <- pca$scores[, 1]
backdrop()
gwpca.check.components 31

points(Data.1$XCOO[pc1 > 0], Data.1$YCOO[pc1 > 0], pch = 16, col = "blue")
points(Data.1$XCOO[pc1 < 0], Data.1$YCOO[pc1 < 0], pch = 16, col = "red")

#Geographically Weighted PCA and mapping the local loadings

# Coordinates of the sites
Coords1 <- as.matrix(cbind(Data.1$XCOO,Data.1$YCOO))
d1s <- SpatialPointsDataFrame(Coords1,as.data.frame(Data.1.scaled))
pca.gw <- gwpca(d1s,vars=colnames(d1s@data),bw=1000000,k=10)
local.loadings <- pca.gw$loadings[, , 1]

# Mapping the winning variable with the highest absolute loading

# note first component only - would need to explore all components..

lead.item <- colnames(local.loadings)[max.col(abs(local.loadings))]

df1p = SpatialPointsDataFrame(Coords1, data.frame(lead = lead.item))
backdrop()
colour <- brewer.pal(8, "Dark2")[match(df1p$lead, unique(df1p$lead))]
plot(df1p, pch = 18, col = colour, add = TRUE)
legend("topleft", as.character(unique(df1p$lead)), pch = 18, col =
brewer.pal(8, "Dark2"))
backdrop()

#Glyph plots give a view of all the local loadings together

glyph.plot(local.loadings, Coords1, add = TRUE)

#it is not immediately clear how to interpret the glyphs fully,

#so inter-actively identify the full loading information using:
check.components(local.loadings, Coords1)

# GWPCA with an optimal bandwidth

bw.choice <- bw.gwpca(d1s,vars=colnames(d1s@data),k=2)
pca.gw.auto <- gwpca(d1s,vars=colnames(d1s@data),bw=bw.choice,k=2)
# note first component only - would need to explore all components..
local.loadings <- pca.gw.auto$loadings[, , 1]

lead.item <- colnames(local.loadings)[max.col(abs(local.loadings))]

# GWPCPLOT for investigating the raw multivariate data

gw.pcplot(d1s, vars=colnames(d1s@data),focus=359, bw = bw.choice)
}

## End(Not run)
32 gwpca.cv

gwpca.check.components
Interaction tool with the GWPCA glyph map

Description
The function interacts with the multivariate glyph plot of GWPCA loadings.

Usage
gwpca.check.components(ld,loc)

Arguments
ld GWPCA loadings returned by gwpca
loc a 2-column numeric array of GWPCA evaluation locations

Note
The function “check.components” (in the early versions of GWmodel) has been renamed as “gw-
pca.check.components”, while the old name is still kept valid.

Author(s)
Binbin Lu <[email protected]>

gwpca.cv Cross-validation score for a specified bandwidth for GWPCA

Description
This function finds the cross-validation score for a specified bandwidth for basic or robust GWPCA.
It can be used to construct the bandwidth function across all possible bandwidths and compared to
that found automatically.

Usage
gwpca.cv(bw,x,loc,k=2,robust=FALSE,kernel="bisquare",adaptive=FALSE,p=2,
theta=0, longlat=F,dMat)
gwpca.cv.contrib 33

Arguments
bw bandwidth used in the weighting function;fixed (distance) or adaptive band-
width(number of nearest neighbours)
x the variable matrix
loc a two-column numeric array of observation coordinates
k the number of retained components; k must be less than the number of variables
robust if TRUE, robust GWPCA will be applied; otherwise basic GWPCA will be ap-
plied
kernel function chosen as follows:
gaussian: wgt = exp(-.5*(vdist/bw)^2);
exponential: wgt = exp(-vdist/bw);
bisquare: wgt = (1-(vdist/bw)^2)^2 if vdist < bw, wgt=0 otherwise;
tricube: wgt = (1-(vdist/bw)^3)^3 if vdist < bw, wgt=0 otherwise;
boxcar: wgt=1 if dist < bw, wgt=0 otherwise
adaptive if TRUE calculate an adaptive kernel where the bandwidth (bw) corresponds to
the number of nearest neighbours (i.e. adaptive distance); default is FALSE,
where a fixed kernel is found (bandwidth is a fixed distance)
p the power of the Minkowski distance, default is 2, i.e. the Euclidean distance
theta an angle in radians to rotate the coordinate system, default is 0
longlat if TRUE, great circle distances will be calculated
dMat a pre-specified distance matrix, it can be calculated by the function gw.dist

Value
CV.score cross-validation score

Author(s)
Binbin Lu <[email protected]>

gwpca.cv.contrib Cross-validation data at each observation location for a GWPCA

Description
This function finds the individual cross-validation score at each observation location, for a GWPCA
model, for a specified bandwidth. These data can be mapped to detect unusually high or low cross-
validations scores.

Usage
gwpca.cv.contrib(x,loc,bw, k=2,robust=FALSE,kernel="bisquare",adaptive=FALSE,
p=2, theta=0, longlat=F,dMat)
34 gwpca.glyph.plot

Arguments
x the variable matrix
loc a two-column numeric array of observation coordinates
bw bandwidth used in the weighting function;fixed (distance) or adaptive band-
width(number of nearest neighbours)
k the number of retained components; k must be less than the number of variables
robust if TRUE, robust GWPCA will be applied; otherwise basic GWPCA will be ap-
plied
kernel function chosen as follows:
gaussian: wgt = exp(-.5*(vdist/bw)^2);
exponential: wgt = exp(-vdist/bw);
bisquare: wgt = (1-(vdist/bw)^2)^2 if vdist < bw, wgt=0 otherwise;
tricube: wgt = (1-(vdist/bw)^3)^3 if vdist < bw, wgt=0 otherwise;
boxcar: wgt=1 if dist < bw, wgt=0 otherwise
adaptive if TRUE calculate an adaptive kernel where the bandwidth (bw) corresponds to
the number of nearest neighbours (i.e. adaptive distance); default is FALSE,
where a fixed kernel is found (bandwidth is a fixed distance)
p the power of the Minkowski distance, default is 2, i.e. the Euclidean distance
theta an angle in radians to rotate the coordinate system, default is 0
longlat if TRUE, great circle distances will be calculated
dMat a pre-specified distance matrix, it can be calculated by the function gw.dist

Value
CV a data vector consisting of squared residuals, whose sum is the cross-validation
score for the specified bandwidth (bw) and component (k).

Author(s)
Binbin Lu <[email protected]>

gwpca.glyph.plot Multivariate glyph plots of GWPCA loadings

Description
This function provides a multivariate glyph plot of GWPCA loadings at each output location.

Usage
gwpca.glyph.plot(ld,loc, r1=50, add=FALSE,alpha=1,sep.contrasts=FALSE)
gwpca.montecarlo.1 35

Arguments
ld GWPCA loadings returned by gwpca
loc a two-column numeric array for providing evaluation locations of GWPCA cal-
ibration
r1 argument for the size of the glyphs, default is 50; glyphs get larger as r1 is
reduced
add if TRUE, add the plot to the existing window.
alpha the level of transparency of glyph from function rgb() and ranges from 0 to max
(fully transparent to opaque)
sep.contrasts allows different types of glyphs and relates to whether absolute loadings are
used (TRUE) or not

Note
The function “glyph.plot” (in the early versions of GWmodel) has been renamed as “gwpca.glyph.plot”,
while the old name is still kept valid.

References
Harris P, Brunsdon C, Charlton M (2011) Geographically weighted principal components analysis.
International Journal of Geographical Information Science 25:1717-1736

gwpca.montecarlo.1 Monte Carlo (randomisation) test for significance of GWPCA eigen-

value variability for the first component only - option 1

Description
This function implements a Monte Carlo (randomisation) test for a basic or robust GW PCA with
the bandwidth pre-specified and constant. The test evaluates whether the GW eigenvalues vary
significantly across space for the first component only.

Usage
gwpca.montecarlo.1(data, bw, vars, k = 2, nsims=99,robust = FALSE, kernel = "bisquare",
adaptive = FALSE, p = 2, theta = 0, longlat = F, dMat)
## S3 method for class 'mcsims'
plot(x, sname="SD of local eigenvalues from randomisations", ...)

Arguments
data a Spatial*DataFrame, i.e. SpatialPointsDataFrame or SpatialPolygonsDataFrame
as defined in package sp
bw bandwidth used in the weighting function, possibly calculated by bw.gwpca;fixed
(distance) or adaptive bandwidth(number of nearest neighbours)
36 gwpca.montecarlo.1

vars a vector of variable names to be evaluated

k the number of retained components; k must be less than the number of variables
nsims the number of simulations for MontCarlo test
robust if TRUE, robust GWPCA will be applied; otherwise basic GWPCA will be ap-
plied
kernel function chosen as follows:
gaussian: wgt = exp(-.5*(vdist/bw)^2);
exponential: wgt = exp(-vdist/bw);
bisquare: wgt = (1-(vdist/bw)^2)^2 if vdist < bw, wgt=0 otherwise;
tricube: wgt = (1-(vdist/bw)^3)^3 if vdist < bw, wgt=0 otherwise;
boxcar: wgt=1 if dist < bw, wgt=0 otherwise
adaptive if TRUE calculate an adaptive kernel where the bandwidth (bw) corresponds to
the number of nearest neighbours (i.e. adaptive distance); default is FALSE,
where a fixed kernel is found (bandwidth is a fixed distance)
p the power of the Minkowski distance, default is 2, i.e. the Euclidean distance
theta an angle in radians to rotate the coordinate system, default is 0
longlat if TRUE, great circle distances will be calculated
dMat a pre-specified distance matrix, it can be calculated by the function gw.dist
x an object of class “mcsims”, returned by the function gwpca.montecarlo.1 or
gwpca.montecarlo.2
sname the label for the observed value on the plot
... arguments passed through (unused)

Value
A list of components:

actual the observed standard deviations (SD) of eigenvalues

sims a vector of the simulated SDs of eigenvalues

Note
The function “montecarlo.gwpca.1” (in the early versions of GWmodel) has been renamed as “gw-
pca.montecarlo.1”, while the old name is still kept valid.

Author(s)
Binbin Lu <[email protected]>

References
Harris P, Brunsdon C, Charlton M (2011) Geographically weighted principal components analysis.
International Journal of Geographical Information Science 25:1717-1736
gwpca.montecarlo.2 37

Examples
## Not run:
data(DubVoter)
DM<-gw.dist(dp.locat=coordinates(Dub.voter))
gmc.res<-gwpca.montecarlo.1(data=Dub.voter, vars=c("DiffAdd", "LARent",
"SC1", "Unempl", "LowEduc"), bw=20,dMat=DM,adaptive=TRUE)
gmc.res
plot(gmc.res)

## End(Not run)

gwpca.montecarlo.2 Monte Carlo (randomisation) test for significance of GWPCA eigen-

value variability for the first component only - option 2

Description
This function implements a Monte Carlo (randomisation) test for a basic or robust GW PCA with the
bandwidth automatically re-selected via the cross-validation approach. The test evaluates whether
the GW eigenvalues vary significantly across space for the first component only.

Usage
gwpca.montecarlo.2(data, vars, k = 2, nsims=99,robust = FALSE, kernel = "bisquare",
adaptive = FALSE, p = 2, theta = 0, longlat = F, dMat)

Arguments
data a Spatial*DataFrame, i.e. SpatialPointsDataFrame or SpatialPolygonsDataFrame
as defined in package sp
vars a vector of variable names to be evaluated
k the number of retained components; k must be less than the number of variables
nsims the number of simulations for MontCarlo test
robust if TRUE, robust GWPCA will be applied; otherwise basic GWPCA will be ap-
plied
kernel function chosen as follows:
gaussian: wgt = exp(-.5*(vdist/bw)^2);
exponential: wgt = exp(-vdist/bw);
bisquare: wgt = (1-(vdist/bw)^2)^2 if vdist < bw, wgt=0 otherwise;
tricube: wgt = (1-(vdist/bw)^3)^3 if vdist < bw, wgt=0 otherwise;
boxcar: wgt=1 if dist < bw, wgt=0 otherwise
adaptive if TRUE calculate an adaptive kernel where the bandwidth (bw) corresponds to
the number of nearest neighbours (i.e. adaptive distance); default is FALSE,
where a fixed kernel is found (bandwidth is a fixed distance)
38 gwr.basic

Value
A list of components:

actual the observed standard deviations (SD) of eigenvalues

sims a vector of the simulated SDs of eigenvalues

Note
The function “montecarlo.gwpca.2” (in the early versions of GWmodel) has been renamed as “gw-
pca.montecarlo.2”, while the old name is still kept valid.

Author(s)
Binbin Lu <[email protected]>

References
Harris P, Brunsdon C, Charlton M (2011) Geographically weighted principal components analysis.
International Journal of Geographical Information Science 25:1717-1736

Examples
## Not run:
data(DubVoter)
DM<-gw.dist(dp.locat=coordinates(Dub.voter))
gmc.res.autow<-gwpca.montecarlo.2(data=Dub.voter, vars=c("DiffAdd", "LARent",
"SC1", "Unempl", "LowEduc"), dMat=DM,adaptive=TRUE)
gmc.res.autow
plot.mcsims(gmc.res.autow)

## End(Not run)

gwr.basic Basic GWR model

Description
This function implements basic GWR
gwr.basic 39

Usage
gwr.basic(formula, data, regression.points, bw, kernel="bisquare",
adaptive=FALSE, p=2, theta=0, longlat=F,dMat,F123.test=F,cv=F, W.vect=NULL)
## S3 method for class 'gwrm'
print(x, ...)

Arguments
formula Regression model formula of a formula object
data a Spatial*DataFrame, i.e. SpatialPointsDataFrame or SpatialPolygonsDataFrame
as defined in package sp
regression.points
a Spatial*DataFrame object, i.e. SpatialPointsDataFrame or SpatialPolygons-
DataFrame as defined in package sp; Note that no diagnostic information will
returned if it is assigned
bw bandwidth used in the weighting function, possibly calculated by bw.gwr;fixed
(distance) or adaptive bandwidth(number of nearest neighbours)
kernel function chosen as follows:
gaussian: wgt = exp(-.5*(vdist/bw)^2);
exponential: wgt = exp(-vdist/bw);
bisquare: wgt = (1-(vdist/bw)^2)^2 if vdist < bw, wgt=0 otherwise;
tricube: wgt = (1-(vdist/bw)^3)^3 if vdist < bw, wgt=0 otherwise;
boxcar: wgt=1 if dist < bw, wgt=0 otherwise
adaptive if TRUE calculate an adaptive kernel where the bandwidth (bw) corresponds to
the number of nearest neighbours (i.e. adaptive distance); default is FALSE,
where a fixed kernel is found (bandwidth is a fixed distance)
p the power of the Minkowski distance, default is 2, i.e. the Euclidean distance
theta an angle in radians to rotate the coordinate system, default is 0
longlat if TRUE, great circle distances will be calculated
dMat a pre-specified distance matrix, it can be calculated by the function gw.dist
F123.test If TRUE, conduct three seperate F-tests according to Leung et al. (2000).
cv if TRUE, cross-validation data will be calculated and returned in the output Spa-
tial*DataFrame
W.vect default NULL, if given it will be used to weight the distance weighting matrix
x an object of class “gwrm”, returned by the function gwr.basic
... arguments passed through (unused)

Value
A list of class “gwrm”:

GW.arguments a list class object including the model fitting parameters for generating the report
file
40 gwr.basic

GW.diagnostic a list class object including the diagnostic information of the model fitting
lm an object of class inheriting from “lm”, see lm.
SDF a SpatialPointsDataFrame (may be gridded) or SpatialPolygonsDataFrame ob-
ject (see package “sp”) integrated with fit.points,GWR coefficient estimates, y
value,predicted values, coefficient standard errors and t-values in its "data" slot.
timings starting and ending time.
this.call the function call used.
Ftest.res results of Leung’s F tests when F123.test is TRUE.

Author(s)
Binbin Lu <[email protected]>

References
Brunsdon, C, Fotheringham, S, Charlton, M (1996), Geographically Weighted Regression: A Method
for Exploring Spatial Nonstationarity. Geographical Analysis 28(4):281-298
Charlton, M, Fotheringham, S, and Brunsdon, C (2007), GWR3.0, http://gwr.nuim.ie/.
Fotheringham S, Brunsdon, C, and Charlton, M (2002), Geographically Weighted Regression: The
Analysis of Spatially Varying Relationships, Chichester: Wiley.
Leung, Y, Mei, CL, and Zhang, WX (2000), Statistical tests for spatial nonstationarity based on the
geographically weighted regression model. Environment and Planning A, 32, 9-32.
Lu, B, Charlton, M, Harris, P, Fotheringham, AS (2014) Geographically weighted regression with
a non-Euclidean distance metric: a case study using hedonic house price data. International Journal
of Geographical Information Science 28(4): 660-681

Examples
data(LondonHP)
DM<-gw.dist(dp.locat=coordinates(londonhp))
##Compare the time consumed with and without a specified distance matrix
## Not run:
system.time(gwr.res<-gwr.basic(PURCHASE~FLOORSZ, data=londonhp, bw=1000,
kernel = "gaussian"))
system.time(DM<-gw.dist(dp.locat=coordinates(londonhp)))
system.time(gwr.res<-gwr.basic(PURCHASE~FLOORSZ, data=londonhp, bw=1000,
kernel = "gaussian", dMat=DM))

## specify an optimum bandwidth by cross-validation appraoch

bw1<-bw.gwr(PURCHASE~FLOORSZ, data=londonhp, kernel = "gaussian",dMat=DM)
gwr.res1<-gwr.basic(PURCHASE~FLOORSZ, data=londonhp, bw=bw1,kernel = "gaussian",
dMat=DM)
gwr.res1
## End(Not run)
data(LondonBorough)

nsa = list("SpatialPolygonsRescale", layout.north.arrow(), offset = c(561900,200900),

scale = 500, col=1)
gwr.bootstrap 41

## Not run:
if(require("RColorBrewer"))
{
mypalette<-brewer.pal(6,"Spectral")
x11()
spplot(gwr.res1$SDF, "FLOORSZ", key.space = "right", cex=1.5, cuts=10,
ylim=c(155840.8,200933.9), xlim=c(503568.2,561957.5),
main="GWR estimated coefficients for FLOORSZ with a fixed bandwidth",
col.regions=mypalette, sp.layout=list(nsa, londonborough))}

## End(Not run)
## Not run:
bw2<-bw.gwr(PURCHASE~FLOORSZ,approach="aic",adaptive=TRUE, data=londonhp,
kernel = "gaussian", dMat=DM)
gwr.res2<-gwr.basic(PURCHASE~FLOORSZ, data=londonhp, bw=bw2,adaptive=TRUE,
kernel = "gaussian", dMat=DM)
gwr.res2
if(require("RColorBrewer"))
{
x11()
spplot(gwr.res2$SDF, "FLOORSZ", key.space = "right", cex=1.5, cuts=10,
ylim=c(155840.8,200933.9), xlim=c(503568.2,561957.5),
main="GWR estimated coefficients for FLOORSZ with an adaptive bandwidth",
col.regions=mypalette, sp.layout=list(nsa,londonborough))}

## End(Not run)

gwr.bootstrap Bootstrap GWR

Description
This function implements bootstrap methods to test for coefficient variability found from GWR
under model assumptions for each of four null hypotheses: MLR, ERR, SMA and LAG models.
Global test statistic results are found, as well local observation-specific test results that can be
mapped.

Usage
gwr.bootstrap(formula, data, kernel="bisquare",approach="AIC", R=99,k.nearneigh=4,
adaptive=FALSE, p=2, theta=0, longlat=FALSE,dMat,verbose=FALSE)
## S3 method for class 'gwrbsm'
print(x, ...)

Arguments
formula Regression model formula of a formula object
data a Spatial*DataFrame, i.e. SpatialPointsDataFrame or SpatialPolygonsDataFrame
as defined in package sp
42 gwr.bootstrap

kernel function chosen as follows:

gaussian: wgt = exp(-.5*(vdist/bw)^2);
exponential: wgt = exp(-vdist/bw);
bisquare: wgt = (1-(vdist/bw)^2)^2 if vdist < bw, wgt=0 otherwise;
tricube: wgt = (1-(vdist/bw)^3)^3 if vdist < bw, wgt=0 otherwise;
boxcar: wgt=1 if dist < bw, wgt=0 otherwise
approach specified by CV for cross-validation approach or by AIC corrected (AICc) ap-
proach
R number of random samples reapted in the bootstrap procedure
k.nearneigh number of nearest neighbours concerned in calbrating ERR, SMA and LAG
models
adaptive if TRUE calculate an adaptive kernel where the bandwidth (bw) corresponds to
the number of nearest neighbours (i.e. adaptive distance); default is FALSE,
where a fixed kernel is found (bandwidth is a fixed distance)
p the power of the Minkowski distance, default is 2, i.e. the Euclidean distance
theta an angle in radians to rotate the coordinate system, default is 0
longlat if TRUE, great circle distances will be calculated
dMat a pre-specified distance matrix, it can be calculated by the function gw.dist
verbose if TRUE and bandwidth selection is undertaken, the bandwidth searches are
reported
x an object of class “gwrbsm”, returned by the function gwr.bootstrap
... arguments passed through (unused)

Value
A list of class “gwrbsm”:
formula Regression model formula of a formula object
results modified statistics reported from comparisons between GWR and MLR, ERR,
SMA and LAG
SDF a SpatialPointsDataFrame (may be gridded) or SpatialPolygonsDataFrame ob-
ject (see package “sp”) integrated with fit.points,GWR coefficient estimates, y
value,predicted values, coefficient standard errors and bootstrap p-values in its
“data” slot.
timings starting and ending time.
this.call the function call used.

Note
This function implements the bootstrap methods introduced in Harris et al. (2017). It provides a
global test statistic (the modified one given in Harris et al. 2017) and a complementary localised
version that can be mapped. The bootstrap methods test for coefficient variability found from GWR
under model assumptions for each of four null hypotheses: i) multiple linear regression model
(MLR); ii) simultaneous autoregressive error model (ERR); iii) moving average error model (SMA)
and iv) simultaneous autoregressive lag model (LAG).
gwr.bootstrap 43

Author(s)
Binbin Lu <[email protected]>

References
Harris, P., Brunsdon, C., Lu, B., Nakaya, T., & Charlton, M. (2017). Introducing bootstrap methods
to investigate coefficient non-stationarity in spatial regression models. Spatial Statistics, 21, 241-
261.

Examples
## Not run:
#Example with the Georgia educational attainment data
data(Georgia)
data(GeorgiaCounties)
coords <- cbind(Gedu.df$X, Gedu.df$Y)
Gedu.spdf <- SpatialPointsDataFrame(coords, Gedu.df)
#Make a SpatialPolygonDataFrame
require(RColorBrewer)
gSRDF <- SpatialPolygonsDataFrame(polygons(Gedu.counties), over(Gedu.counties,
Gedu.spdf),match.ID=T)
mypalette.1 <- brewer.pal(11,"Spectral")
X11(width=9,height=8)
spplot(gSRDF, names(gSRDF)[c(5,7:9)], col.regions=mypalette.1,
cuts=10, par.settings=list(fontsize=list(text=15)),
main=expression(paste("Georgia educational attainment predictor data")))
bsm.res <- gwr.bootstrap(PctBach~PctRural+PctEld+PctFB+PctPov, gSRDF,
R=999, longlat=T)
bsm.res
#local bootstrap tests with respect to: MLR, ERR, SMA and LAG models.
mypalette.local.test <- brewer.pal(10,"Spectral")
X11(width=12,height=16)
spplot(bsm.res$SDF, names(bsm.res$SDF)[14:17], col.regions=mypalette.local.test,
cuts=9, par.settings=list(fontsize=list(text=15)),
main=expression(paste("Local p-values for each coefficient of the MLR model
null hypothesis")))

X11(width=12,height=16)
spplot(bsm.res$SDF, names(bsm.res$SDF)[19:22], col.regions=mypalette.local.test,
cuts=9, par.settings=list(fontsize=list(text=15)),
main=expression(paste("Local p-values for each coefficient of the ERR model
null hypothesis")))
X11(width=12,height=16)
spplot(bsm.res$SDF, names(bsm.res$SDF)[24:27], col.regions=mypalette.local.test,
cuts=9, par.settings=list(fontsize=list(text=15)),
main=expression(paste("Local p-values for each coefficient of the SMA model null
hypothesis")))

X11(width=12,height=16)
spplot(bsm.res$SDF, names(bsm.res$SDF)[29:32], col.regions=mypalette.local.test,
cuts=9, par.settings=list(fontsize=list(text=15)),
main=expression(paste("Local p-values for each coefficient of the LAG model null
44 gwr.collin.diagno

hypothesis")))
################################################################################
#Example with Dublin voter data
data(DubVoter)
X11(width=9,height=8)
spplot(Dub.voter, names(Dub.voter)[c(5,7,9,10)], col.regions=mypalette.1,
cuts=10, par.settings=list(fontsize=list(text=15)),
main=expression(paste("Dublin voter turnout predictor data")))
bsm.res1 <- gwr.bootstrap(GenEl2004~LARent+Unempl+Age18_24+Age25_44, Dub.voter
, R=999)
bsm.res1

#local bootstrap tests with respect to: MLR, ERR, SMA and LAG models.
X11(width=11,height=8)
spplot(bsm.res1$SDF, names(bsm.res1$SDF)[14:17], col.regions=mypalette.local.test,
cuts=9, par.settings=list(fontsize=list(text=15)),
main=expression(paste("Local p-values for each coefficient of the MLR model null
hypothesis")))
X11(width=11,height=8)
spplot(bsm.res1$SDF, names(bsm.res1$SDF)[19:22], col.regions=mypalette.local.test,
cuts=9, par.settings=list(fontsize=list(text=15)),
main=expression(paste("Local p-values for each coefficient of the ERR model null
hypothesis")))
X11(width=11,height=8)
spplot(bsm.res1$SDF, names(bsm.res1$SDF)[24:27], col.regions=mypalette.local.test,
cuts=9, par.settings=list(fontsize=list(text=15)),
main=expression(paste("Local p-values for each coefficient of the SMA model
null hypothesis")))
X11(width=11,height=8)
spplot(bsm.res1$SDF, names(bsm.res1$SDF)[29:32], col.regions=mypalette.local.test,
cuts=9, par.settings=list(fontsize=list(text=15)),
main=expression(paste("Local p-values for each coefficient of the LAG model
null hypothesis")))

## End(Not run)

gwr.collin.diagno Local collinearity diagnostics for basic GWR

Description
This function provides a series of local collinearity diagnostics for the independent variables of a
basic GWR model.

Usage
gwr.collin.diagno(formula, data, bw, kernel="bisquare",
adaptive=FALSE, p=2, theta=0, longlat=F,dMat)
gwr.collin.diagno 45

Value
corr.mat Local correlation matrix
VIF Local Variance inflation factors (VIFs) matrix
local_CN Local condition numbers
VDP Local variance-decomposition proportions
SDF a SpatialPointsDataFrame (may be gridded) or SpatialPolygonsDataFrame ob-
ject (see package “sp”) integrated with VIF, local_CN, VDP and corr.mat

Author(s)
Binbin Lu <[email protected]>

References
Wheeler D, Tiefelsdorf M (2005) Multicollinearity and correlation among local regression coeffi-
cients in geographically weighted regression. Journal of Geographical Systems 7:161-187
Wheeler D (2007) Diagnostic tools and a remedial method for collinearity in geographically weighted
regression. Environment and Planning A 39:2464-2481
Gollini I, Lu B, Charlton M, Brunsdon C, Harris P (2015) GWmodel: an R Package for explor-
ing Spatial Heterogeneity using Geographically Weighted Models. Journal of Statistical Software,
63(17):1-50
46 gwr.cv

gwr.cv Cross-validation score for a specified bandwidth for basic GWR

Description
This function finds the cross-validation score for a specified bandwidth for basic GWR. It can be
used to construct the bandwidth function across all possible bandwidths and compared to that found
automatically.

Usage
gwr.cv(bw, X, Y, kernel="bisquare",adaptive=FALSE, dp.locat, p=2, theta=0,
longlat=F,dMat, verbose=T)

Value
CV.score cross-validation score

Author(s)
Binbin Lu <[email protected]>
gwr.cv.contrib 47

gwr.cv.contrib Cross-validation data at each observation location for a basic GWR

model

Description
This function finds the individual cross-validation score at each observation location, for a basic
GWR model, for a specified bandwidth. These data can be mapped to detect unusually high or low
cross-validations scores.

Usage
gwr.cv.contrib(bw, X, Y, kernel="bisquare",adaptive=FALSE, dp.locat, p=2,
theta=0, longlat=F,dMat)

Value
CV a data vector consisting of squared residuals, whose sum is the cross-validation
score for the specified bandwidth.

Author(s)
Binbin Lu <[email protected]>
48 gwr.hetero

gwr.hetero Heteroskedastic GWR

Description
This function implements a heteroskedastic GWR model

Usage
gwr.hetero(formula, data, regression.points, bw, kernel="bisquare",
adaptive=FALSE, tol=0.0001,maxiter=50,verbose=T,
p=2, theta=0, longlat=F,dMat)

Arguments
formula Regression model formula of a formula object
data a Spatial*DataFrame, i.e. SpatialPointsDataFrame or SpatialPolygonsDataFrame
as defined in package sp
regression.points
a Spatial*DataFrame object, i.e. SpatialPointsDataFrame or SpatialPolygons-
DataFrame as defined in package sp
bw bandwidth used in the weighting function, possibly calculated by bw.gwr;fixed
(distance) or adaptive bandwidth(number of nearest neighbours)
kernel function chosen as follows:
gaussian: wgt = exp(-.5*(vdist/bw)^2);
exponential: wgt = exp(-vdist/bw);
bisquare: wgt = (1-(vdist/bw)^2)^2 if vdist < bw, wgt=0 otherwise;
tricube: wgt = (1-(vdist/bw)^3)^3 if vdist < bw, wgt=0 otherwise;
boxcar: wgt=1 if dist < bw, wgt=0 otherwise
adaptive if TRUE calculate an adaptive kernel where the bandwidth (bw) corresponds to
the number of nearest neighbours (i.e. adaptive distance); default is FALSE,
where a fixed kernel is found (bandwidth is a fixed distance)
tol the threshold that determines the convergence of the iterative procedure
maxiter the maximum number of times to try the iterative procedure
verbose logical, if TRUE verbose output will be made from the iterative procedure
p the power of the Minkowski distance, default is 2, i.e. the Euclidean distance
theta an angle in radians to rotate the coordinate system, default is 0
longlat if TRUE, great circle distances will be calculated
dMat a pre-specified distance matrix, it can be calculated by the function gw.dist

Value
SDF a SpatialPointsDataFrame (may be gridded) or SpatialPolygonsDataFrame ob-
ject (see package “sp”) integrated with coefficient estimates in its "data" slot.
gwr.lcr 49

Author(s)
Binbin Lu <[email protected]>

References
Fotheringham S, Brunsdon, C, and Charlton, M (2002), Geographically Weighted Regression: The
Analysis of Spatially Varying Relationships, Chichester: Wiley.
Harris P, Fotheringham AS, Juggins S (2010) Robust geographically weighed regression: a tech-
nique for quantifying spatial relationships between freshwater acidification critical loads and catch-
ment attributes. Annals of the Association of American Geographers 100(2): 286-306
Harris P, Brunsdon C, Fotheringham AS (2011) Links, comparisons and extensions of the geo-
graphically weighted regression model when used as a spatial predictor. Stochastic Environmental
Research and Risk Assessment 25:123-138

gwr.lcr GWR with a locally-compensated ridge term

Description
To address possible local collinearity problems in basic GWR, GWR-LCR finds local ridge parame-
ters at affected locations (set by a user-specified threshold for the design matrix condition number).

Usage
gwr.lcr(formula, data, regression.points, bw, kernel="bisquare",
lambda=0,lambda.adjust=FALSE,cn.thresh=NA,
adaptive=FALSE, p=2, theta=0, longlat=F,cv=T,dMat)
## S3 method for class 'gwrlcr'
print(x, ...)

p the power of the Minkowski distance, default is 2, i.e. the Euclidean distance
lambda option for a globally-defined (constant) ridge parameter. Default is lambda=0,
which gives a basic GWR fit
lambda.adjust a locally-varying ridge parameter. Default FALSE, refers to: (i) a basic GWR
without a local ridge adjustment (i.e. lambda=0, everywhere); or (ii) a penalised
GWR with a global ridge adjustment (i.e. lambda is user-specified as some
constant, other than 0 everywhere); if TRUE, use cn.tresh to set the maximum
condition number. Here for locations with a condition number (for its local
design matrix) above this user-specified threshold, a local ridge parameter is
found
cn.thresh maximum value for condition number, commonly set between 20 and 30
adaptive if TRUE calculate an adaptive kernel where the bandwidth (bw) corresponds to
the number of nearest neighbours (i.e. adaptive distance); default is FALSE,
where a fixed kernel is found (bandwidth is a fixed distance)
theta an angle in radians to rotate the coordinate system, default is 0
longlat if TRUE, great circle distances will be calculated
cv if TRUE, ’cross-validation data will be calculated and returned in the output
Spatial*DataFrame
dMat a pre-specified distance matrix, it can be calculated by the function gw.dist
x an object of class “gwrlcr”, returned by the function gwr.lcr
... arguments passed through (unused)

Value
A list of class “rgwr”:

SDF a SpatialPointsDataFrame (may be gridded) or SpatialPolygonsDataFrame ob-

ject (see package "sp") with coordinates of regression.points in its "data" slot.
GW.arguments parameters used for the LCR-GWR calibration
GW.diagnostic diagnostic information is given when data points are also used as regression
locations
timings timing information for running this function
this.call the function call used.

Author(s)
Binbin Lu <[email protected]>

References
Wheeler D (2007) Diagnostic tools and a remedial method for collinearity in geographically weighted
regression. Environment and Planning A 39:2464-2481
Brunsdon C, Charlton M, Harris P (2012) Living with collinearity in Local Regression Models.
GISRUK 2012, Lancaster, UK
gwr.lcr.cv 51

Brunsdon C, Charlton M, Harris P (2012) Living with collinearity in Local Regression Models.
Spatial Accuracy 2012, Brazil
Gollini I, Lu B, Charlton M, Brunsdon C, Harris P (2015) GWmodel: an R Package for explor-
ing Spatial Heterogeneity using Geographically Weighted Models. Journal of Statistical Software
63(17): 1-50

Examples

data(DubVoter)
require(RColorBrewer)

# Function to find the global condition number (CN)

BKW_cn <- function (X) {
p <- dim(X)[2]
Xscale <- sweep(X, 2, sqrt(colSums(X^2)), "/")
Xsvd <- svd(Xscale)$d
cn <- Xsvd[1] / Xsvd[p]
cn
}
#
X <- cbind(1,Dub.voter@data[,3:10])
head(X)
CN.global <- BKW_cn(X)
CN.global
## Not run:
# gwr.lcr function with a global bandwidth to check that the global CN is found
gwr.lcr1 <- gwr.lcr(GenEl2004~DiffAdd+LARent+SC1+Unempl+LowEduc+Age18_24
+Age25_44+Age45_64, data=Dub.voter, bw=10000000000)
summary(gwr.lcr1$SDF$Local_CN)

# Find and map the local CNs from a basic GWR fit using the lcr-gwr function
#(note this is NOT the locally-compensated ridge GWR fit as would need to set
#lambda.adjust=TRUE and cn.thresh=30, say)

bw.lcr2 <- bw.gwr.lcr(GenEl2004~DiffAdd+LARent+SC1+Unempl+LowEduc+Age18_24

+Age25_44+Age45_64, data=Dub.voter, kernel="bisquare", adaptive=TRUE)
gwr.lcr2 <- gwr.lcr(GenEl2004~DiffAdd+LARent+SC1+Unempl+LowEduc+Age18_24
+Age25_44+Age45_64, data=Dub.voter, bw=bw.lcr2, kernel="bisquare", adaptive=TRUE)
if(require("RColorBrewer"))
spplot(gwr.lcr2$SDF,"Local_CN",col.regions=brewer.pal(9,"YlOrRd"),cuts=8,
main="Local CN")

## End(Not run)

gwr.lcr.cv Cross-validation score for a specified bandwidth for GWR-LCR model

52 gwr.lcr.cv

Description
This function finds the cross-validation score for a specified bandwidth for GWR-LCR. It can be
used to construct the bandwidth function across all possible bandwidths and compared to that found
automatically.

Usage
gwr.lcr.cv(bw,X,Y,locs,kernel="bisquare",
lambda=0,lambda.adjust=FALSE,cn.thresh=NA,
adaptive=FALSE, p=2, theta=0, longlat=F,dMat)

Arguments
bw bandwidth used in the weighting function;fixed (distance) or adaptive band-
width(number of nearest neighbours)
X a numeric matrix of the independent data with an extra column of “ones” for the
1st column
Y a column vector of the dependent data
kernel function chosen as follows:
gaussian: wgt = exp(-.5*(vdist/bw)^2);
exponential: wgt = exp(-vdist/bw);
bisquare: wgt = (1-(vdist/bw)^2)^2 if vdist < bw, wgt=0 otherwise;
tricube: wgt = (1-(vdist/bw)^3)^3 if vdist < bw, wgt=0 otherwise;
boxcar: wgt=1 if dist < bw, wgt=0 otherwise
locs a two-column numeric array of observation coordinates
lambda option for a globally-defined (constant) ridge parameter. Default is lambda=0,
which gives a basic GWR fit
lambda.adjust a locally-varying ridge parameter. Default FALSE, refers to: (i) a basic GWR
without a local ridge adjustment (i.e. lambda=0, everywhere); or (ii) a penalised
GWR with a global ridge adjustment (i.e. lambda is user-specified as some
constant, other than 0 everywhere); if TRUE, use cn.tresh to set the maximum
condition number. Here for locations with a condition number (for its local
design matrix) above this user-specified threshold, a local ridge parameter is
found
cn.thresh maximum value for condition number, commonly set between 20 and 30
adaptive if TRUE calculate an adaptive kernel where the bandwidth (bw) corresponds to
the number of nearest neighbours (i.e. adaptive distance); default is FALSE,
where a fixed kernel is found (bandwidth is a fixed distance)
p the power of the Minkowski distance, default is 2, i.e. the Euclidean distance
theta an angle in radians to rotate the coordinate system, default is 0
longlat if TRUE, great circle distances will be calculated
dMat a pre-specified distance matrix, it can be calculated by the function gw.dist
gwr.lcr.cv.contrib 53

Value
CV.score cross-validation score

Author(s)
Binbin Lu <[email protected]>

gwr.lcr.cv.contrib Cross-validation data at each observation location for the GWR-LCR

model

Description
This function finds the individual cross-validation score at each observation location, for a GWR-
LCR model, for a specified bandwidth. These data can be mapped to detect unusually high or low
cross-validations scores.

Usage
gwr.lcr.cv.contrib(bw,X,Y,locs,kernel="bisquare",
lambda=0,lambda.adjust=FALSE,cn.thresh=NA,
adaptive=FALSE, p=2, theta=0, longlat=F,dMat)

Arguments
bw bandwidth used in the weighting function;fixed (distance) or adaptive band-
width(number of nearest neighbours)
X a numeric matrix of the independent data with an extra column of “ones” for the
1st column
Y a column vector of the dependent data
locs a two-column numeric array of observation coordinates
kernel function chosen as follows:
gaussian: wgt = exp(-.5*(vdist/bw)^2);
exponential: wgt = exp(-vdist/bw);
bisquare: wgt = (1-(vdist/bw)^2)^2 if vdist < bw, wgt=0 otherwise;
tricube: wgt = (1-(vdist/bw)^3)^3 if vdist < bw, wgt=0 otherwise;
boxcar: wgt=1 if dist < bw, wgt=0 otherwise
lambda option for a globally-defined (constant) ridge parameter. Default is lambda=0,
which gives a basic GWR fit
lambda.adjust a locally-varying ridge parameter. Default FALSE, refers to: (i) a basic GWR
without a local ridge adjustment (i.e. lambda=0, everywhere); or (ii) a penalised
GWR with a global ridge adjustment (i.e. lambda is user-specified as some
constant, other than 0 everywhere); if TRUE, use cn.tresh to set the maximum
condition number. Here for locations with a condition number (for its local
design matrix) above this user-specified threshold, a local ridge parameter is
found
54 gwr.mink.approach

cn.thresh maximum value for condition number, commonly set between 20 and 30
adaptive if TRUE calculate an adaptive kernel where the bandwidth (bw) corresponds to
the number of nearest neighbours (i.e. adaptive distance); default is FALSE,
where a fixed kernel is found (bandwidth is a fixed distance)
p the power of the Minkowski distance, default is 2, i.e. the Euclidean distance
theta an angle in radians to rotate the coordinate system, default is 0
longlat if TRUE, great circle distances will be calculated
dMat a pre-specified distance matrix, it can be calculated by the function gw.dist

Value
CV a data vector consisting of squared residuals, whose sum is the cross-validation
score for the specified bandwidth.

Author(s)
Binbin Lu <[email protected]>

gwr.mink.approach Minkovski approach for GWR

Description
This function implements the Minkovski approach to select an ’optimum’ distance metric for cali-
brating a GWR model.

Usage
gwr.mink.approach(formula, data, criterion="AIC", bw, bw.sel.approach = "AIC",adaptive=F,
kernel="bisquare", p.vals=seq(from=0.25, to=8, length.out=32), p.inf = T,
theta.vals = seq(from=0, to=0.5*pi, length.out=10), verbose=F,
nlower = 10)

adaptive if TRUE calculate an adaptive kernel where the bandwidth (bw) corresponds to
the number of nearest neighbours (i.e. adaptive distance); default is FALSE,
where a fixed kernel is found (bandwidth is a fixed distance)
kernel function chosen as follows:
gaussian: wgt = exp(-.5*(vdist/bw)^2);
exponential: wgt = exp(-vdist/bw);
bisquare: wgt = (1-(vdist/bw)^2)^2 if vdist < bw, wgt=0 otherwise;
tricube: wgt = (1-(vdist/bw)^3)^3 if vdist < bw, wgt=0 otherwise;
boxcar: wgt=1 if dist < bw, wgt=0 otherwise
p.vals a collection of positive numbers used as the power of the Minkowski distance
p.inf if TRUE, Chebyshev distance is tried for model calibration, i.e. p is infinity
theta.vals a collection of values used as angles in radians to rotate the coordinate system
verbose if TRUE and bandwidth selection is undertaken, the bandwidth searches are
reported
nlower the minmum number of nearest neighbours if an adaptive kernel is used

Value

A list of:

diag.df a data frame with four columns (p, theta, bandwidth, AICc/CV), each row cor-
responds to a calibration
coefs.all a list class object including all the estimated coefficients

Note

The function “mink.approach” (in the early versions of GWmodel) has been renamed as “gwr.mink.approach”,
while the old name is still kept valid.

Author(s)

Binbin Lu <[email protected]>

References

Lu, B, Charlton, M, Brunsdon, C & Harris, P(2016). The Minkowski approach for choosing the
distance metric in Geographically Weighted Regression. International Journal of Geographical In-
formation Science, 30(2): 351-368.
56 gwr.mink.pval

gwr.mink.matrixview Visualisation of the results from gwr.mink.approach

Description

This function visualises the AICc/CV results from the gwr.mink.approach.

Usage

gwr.mink.matrixview(diag.df, znm=colnames(diag.df)[4], criterion="AIC")

Arguments

diag.df the first part of a list object returned by gwr.mink.approach

znm the name of the forth column in diag.df
criterion the criterion used for distance metric selection in gwr.mink.approach

Note

The function “mink.matrixview” (in the early versions of GWmodel) has been renamed as “gwr.mink.matrixview”,
while the old name is still kept valid.

Author(s)

Binbin Lu <[email protected]>

References

gwr.mink.pval Select the values of p for the Minkowski approach for GWR

Description

These functions implement heuristics to select the values of p from two intervals: (0, 2] in a ’back-
ward’ direction and (2, Inf) in a ’forward’ direction.
gwr.mink.pval 57

Usage
gwr.mink.pval(formula, data, criterion="AIC", bw, bw.sel.approach = "AIC",
adaptive=F, kernel="bisquare", left.interval=0.25,
right.interval=0.5,drop.tol=3, theta0=0,verbose=F,nlower = 10)
gwr.mink.pval.forward(formula, data, bw, bw.sel.approach = "AIC",
adaptive=F, kernel="bisquare", p.max=Inf,p.min=2,
interval=0.5,drop.tol=3, theta0=0,verbose=F,nlower = 10)
gwr.mink.pval.backward(formula, data, bw, bw.sel.approach = "AIC",
adaptive=F, kernel="bisquare", p.max=2,p.min=0.1,
interval=0.5,drop.tol=3, theta0=0,verbose=F,nlower = 10)
## S3 method for class 'pvlas'
plot(x, ...)

Arguments
formula Regression model formula of a formula object
data a Spatial*DataFrame, i.e. SpatialPointsDataFrame or SpatialPolygonsDataFrame
as defined in package sp
criterion the criterion used for distance metric selection, AICc ("AICc") or cross-validation
("CV") score; default is "AICc"
bw bandwidth used in the weighting function, possibly calculated by bw.gwr;fixed
(distance) or adaptive bandwidth(number of nearest neighbours)
bw.sel.approach
approach used to seclect an optimum bandwidth for each calibration if no band-
width (bw) is given; specified by CV for cross-validation approach or by AIC
corrected (AICc) approach
adaptive if TRUE calculate an adaptive kernel where the bandwidth (bw) corresponds to
the number of nearest neighbours (i.e. adaptive distance); default is FALSE,
where a fixed kernel is found (bandwidth is a fixed distance)
kernel function chosen as follows:
gaussian: wgt = exp(-.5*(vdist/bw)^2);
exponential: wgt = exp(-vdist/bw);
bisquare: wgt = (1-(vdist/bw)^2)^2 if vdist < bw, wgt=0 otherwise;
tricube: wgt = (1-(vdist/bw)^3)^3 if vdist < bw, wgt=0 otherwise;
boxcar: wgt=1 if dist < bw, wgt=0 otherwise
left.interval the step-size for searching the left interval (0, 2] in a ’backward’ direction
right.interval the step-size for searching the right interval (2, Inf) in a ’forward’ direction
p.max the maximum value of p
p.min the minimum value of p
interval the step-size for searching the given interval in a ’backward’ or ’forward’ direc-
tion
drop.tol an AICc difference threshold to define whether the values of p to be dropped or
not
58 gwr.mixed

theta0 a fixed rotation angle in radians

verbose if TRUE and bandwidth selection is undertaken, the bandwidth searches are
reported
nlower the minmum number of nearest neighbours if an adaptive kernel is used
x an object of class “pvlas”, returned by these functions
... arguments passed through (unused)

Value

A list of:

p.vals a vector of tried values of p

cretion.vals a vector of criterion values (AICc or CV) for tried values of p
p.dropped a vector of boolean to label whether a value of p to be dropped or not: TRUE
means to be dropped and FALSE means to be used for the Minkowski approach

Author(s)

Binbin Lu <[email protected]>

References

gwr.mixed Mixed GWR

Description

This function implements mixed (semiparametric) GWR

Usage

gwr.mixed(formula, data, regression.points, fixed.vars,

intercept.fixed=FALSE, bw, diagnostic=T, kernel="bisquare",
adaptive=FALSE, p=2, theta=0, longlat=F,dMat)
gwr.mixed 59

Arguments
formula Regression model formula of a formula object
data a Spatial*DataFrame, i.e. SpatialPointsDataFrame or SpatialPolygonsDataFrame
as defined in package sp
regression.points
a Spatial*DataFrame object, i.e. SpatialPointsDataFrame or SpatialPolygons-
DataFrame as defined in package sp
fixed.vars independent variables that appeared in the formula that are to be treated as global
intercept.fixed
logical, if TRUE the intercept will be treated as global
bw bandwidth used in the weighting function, possibly calculated by bw.gwr;fixed
(distance) or adaptive bandwidth(number of nearest neighbours)
diagnostic logical, if TRUE the diagnostics will be calculated
kernel function chosen as follows:
gaussian: wgt = exp(-.5*(vdist/bw)^2);
exponential: wgt = exp(-vdist/bw);
bisquare: wgt = (1-(vdist/bw)^2)^2 if vdist < bw, wgt=0 otherwise;
tricube: wgt = (1-(vdist/bw)^3)^3 if vdist < bw, wgt=0 otherwise;
boxcar: wgt=1 if dist < bw, wgt=0 otherwise
adaptive if TRUE calculate an adaptive kernel where the bandwidth (bw) corresponds to
the number of nearest neighbours (i.e. adaptive distance); default is FALSE,
where a fixed kernel is found (bandwidth is a fixed distance)
p the power of the Minkowski distance, default is 2, i.e. the Euclidean distance
theta an angle in radians to rotate the coordinate system, default is 0
longlat if TRUE, great circle distances will be calculated
dMat a pre-specified distance matrix, it can be calculated by the function gw.dist

Value
A list of class “mgwr”:

GW.arguments a list class object including the model fitting parameters for generating the report
file
aic AICc value from this calibration
df.used effective degree of freedom
rss residual sum of squares
SDF a SpatialPointsDataFrame (may be gridded) or SpatialPolygonsDataFrame ob-
ject (see package “sp”) integrated with coefficient estimates in its "data" slot.
timings starting and ending time.
this.call the function call used.
60 gwr.model.selection

Note
For an alternative formulation of mixed GWR, please refer to GWR 4, which provides useful
tools for automatic bandwidth selection. This windows-based software also implements generalised
mixed GWR.

Author(s)
Binbin Lu <[email protected]>

References
Fotheringham S, Brunsdon, C, and Charlton, M (2002), Geographically Weighted Regression: The
Analysis of Spatially Varying Relationships, Chichester: Wiley.
Brunsdon C, Fotheringham AS, Charlton ME (1999) Some notes on parametric signficance tests
for geographically weighted regression. Journal of Regional Science 39(3):497-524
Mei L-M, He S-Y, Fang K-T (2004) A note on the mixed geographically weighted regression model.
Journal of regional science 44(1):143-157
Mei L-M, Wang N, Zhang W-X (2006) Testing the importance of the explanatory variables in a
mixed geographically weighted regression model. Environment and Planning A 38:587-598
Nakaya T, Fotheringham AS, Brunsdon C, Charlton M (2005) Geographically Weighted Poisson
Regression for Disease Association Mapping, Statistics in Medicine 24: 2695-2717
Nakaya T et al. (2011) GWR4.0, http://gwr.nuim.ie/.

gwr.model.selection Model selection for GWR with a given set of independent variables

Description
This function selects one GWR model from many alternatives based on the AICc values.

Usage
gwr.model.selection(DeVar=NULL,InDeVars=NULL, data=list(),bw=NULL,approach="CV",
adaptive=F,kernel="bisquare",dMat=NULL,p=2, theta=0, longlat=F)

Arguments
DeVar dependent variable
InDeVars a vector of independent variables for model selection
data a Spatial*DataFrame, i.e. SpatialPointsDataFrame or SpatialPolygonsDataFrame
as defined in package sp
bw bandwidth used in the weighting function, possibly calculated by bw.gwr
approach specified by CV (cv) for cross validation approach or AIC (aic) for selecting
bandwidth by AICc values
gwr.model.selection 61

adaptive if TRUE calculate an adaptive kernel where the bandwidth (bw) corresponds to
the number of nearest neighbours (i.e. adaptive distance); default is FALSE,
where a fixed kernel is found (bandwidth is a fixed distance)
kernel function chosen as follows:
gaussian: wgt = exp(-.5*(vdist/bw)^2);
exponential: wgt = exp(-vdist/bw);
bisquare: wgt = (1-(vdist/bw)^2)^2 if vdist < bw, wgt=0 otherwise;
tricube: wgt = (1-(vdist/bw)^3)^3 if vdist < bw, wgt=0 otherwise;
boxcar: wgt=1 if dist < bw, wgt=0 otherwise
dMat a pre-specified distance matrix, it can be calculated by the function gw.dist
p the power of the Minkowski distance, default is 2, i.e. the Euclidean distance
theta an angle in radians to rotate the coordinate system, default is 0
longlat if TRUE, great circle distances will be calculated

Value
A list of:

model.list a list of all the tried GWR models consisted of formulas and variables.
GWR.df a data frame consited of four columns: bandwidth, AIC, AICc, RSS

Note
The algorithm for selecting GWR models consists of the following four steps:
Step 1. Start by calibrating all the possible bivariate GWR models by sequentially regressing a
single independent variable against the dependent variable;
Step 2. Find the best performing model which produces the minimum AICc value, and permanently
include the corresponding independent variable in subsequent models;
Step 3. Sequentially introduce a variable from the remaining group of independent variables to
construct new models with the permanently included independent variables, and determine the next
permanently included variable from the best fitting model that has the minimum AICc value;
Step 4. Repeat step 3 until all the independent variables are permanently included in the model.
In this procedure, the independent variables are iteratively included into the model in a "forward"
direction. Note that there is a clear distinction between the different number of involved variables
in a selection step, which can be called model levels.

Author(s)
Binbin Lu <[email protected]>

References
Lu, B, Charlton, M, Harris, P, Fotheringham, AS (2014) Geographically weighted regression with
a non-Euclidean distance metric: a case study using hedonic house price data. International Journal
of Geographical Information Science 28(4): 660-681
62 gwr.model.sort

gwr.model.sort Sort the results of the GWR model selection function

gwr.model.selection.

Description

Sort the results from the GWR model selection function gwr.model.selection

Usage

gwr.model.sort(Sorting.list , numVars, ruler.vector)

Arguments

Sorting.list a list returned by function gwr.model.selection

numVars the number of independent variables involved in model selection
ruler.vector a numeric vector as the sorting basis

Note

The function sorts the results of model selection within individual levels.
The function “model.sort.gwr” (in the early versions of GWmodel) has been renamed as “gwr.model.sort”,
while the old name is still kept valid.

Author(s)

Binbin Lu <[email protected]>

gwr.model.view Visualise the GWR models from gwr.model.selection

Description
This function visualises the GWR models from gwr.model.selection.

Usage
gwr.model.view(DeVar, InDeVars, model.list)

Arguments
DeVar dependent variable
InDeVars a vector of independent variables for model selection
model.list a list of all GWR model tried in gwr.model.selection

Note
The function “model.view.gwr” (in the early versions of GWmodel) has been renamed as “gwr.model.view”,
while the old name is still kept valid.

Author(s)
Binbin Lu <[email protected]>

See Also
gwr.model.selection, gwr.model.sort

Examples
## Not run:
data(LondonHP)
DM<-gw.dist(dp.locat=coordinates(londonhp))
DeVar<-"PURCHASE"
InDeVars<-c("FLOORSZ","GARAGE1","BLDPWW1","BLDPOSTW")
model.sel<-gwr.model.selection(DeVar,InDeVars, data=londonhp,
kernel = "gaussian", dMat=DM,bw=5000)
model.list<-model.sel[[1]]
gwr.model.view(DeVar, InDeVars, model.list=model.list)

## End(Not run)
64 gwr.montecarlo

gwr.montecarlo Monte Carlo (randomisation) test for significance of GWR parameter

variability

Description
This function implements a Monte Carlo (randomisation) test to test for significant (spatial) vari-
ability of a GWR model’s parameters or coefficients.

Usage
gwr.montecarlo(formula, data = list(),nsims=99, kernel="bisquare",adaptive=F, bw,
p=2, theta=0, longlat=F,dMat)

Arguments
formula Regression model formula of a formula object
data a Spatial*DataFrame, i.e. SpatialPointsDataFrame or SpatialPolygonsDataFrame
as defined in package sp
nsims the number of randomisations
kernel function chosen as follows:
gaussian: wgt = exp(-.5*(vdist/bw)^2);
exponential: wgt = exp(-vdist/bw);
bisquare: wgt = (1-(vdist/bw)^2)^2 if vdist < bw, wgt=0 otherwise;
tricube: wgt = (1-(vdist/bw)^3)^3 if vdist < bw, wgt=0 otherwise;
boxcar: wgt=1 if dist < bw, wgt=0 otherwise
adaptive if TRUE calculate an adaptive kernel where the bandwidth (bw) corresponds to
the number of nearest neighbours (i.e. adaptive distance); default is FALSE,
where a fixed kernel is found (bandwidth is a fixed distance)
bw bandwidth used in the weighting function, possibly calculated by bw.gwr
p the power of the Minkowski distance, default is 2, i.e. the Euclidean distance
theta an angle in radians to rotate the coordinate system, default is 0
longlat if TRUE, great circle distances will be calculated
dMat a pre-specified distance matrix, it can be calculated by the function gw.dist

Value
pmat A vector containing p-values for all the GWR parameters

Note
The function “montecarlo.gwr” (in the early versions of GWmodel) has been renamed as “gwr.montecarlo”,
while the old name is still kept valid.
gwr.multiscale 65

Author(s)
Binbin Lu <[email protected]>

References
Brunsdon C, Fotheringham AS, Charlton ME (1998) Geographically weighted regression - mod-
elling spatial non-stationarity. Journal of the Royal Statistical Society, Series D-The Statistician
47(3):431-443
Fotheringham S, Brunsdon, C, and Charlton, M (2002), Geographically Weighted Regression: The
Analysis of Spatially Varying Relationships, Chichester: Wiley.
Charlton, M, Fotheringham, S, and Brunsdon, C (2007), GWR3.0.

Examples
## Not run:
data(LondonHP)
DM<-gw.dist(dp.locat=coordinates(londonhp))
bw<-bw.gwr(PURCHASE~FLOORSZ,data=londonhp,dMat=DM, kernel="gaussian")
#See any difference in the next two commands and why?
res.mont1<-gwr.montecarlo(PURCHASE~PROF+FLOORSZ, data = londonhp,dMat=DM,
nsim=99, kernel="gaussian", adaptive=FALSE, bw=3000)
res.mont2<-gwr.montecarlo(PURCHASE~PROF+FLOORSZ, data = londonhp,dMat=DM,
nsim=99, kernel="gaussian", adaptive=FALSE, bw=300000000000)

## End(Not run)

gwr.multiscale Multiscale GWR

Description
This function implements multiscale GWR to detect variations in regression relationships across
different spatial scales. This function can not only find a different bandwidth for each relationship
but also (and simultaneously) find a different distance metric for each relationship (if required to do
so).

Usage
gwr.multiscale(formula, data, kernel="bisquare", adaptive=FALSE, criterion="dCVR",
max.iterations=2000,threshold=0.00001, dMats, p.vals, theta.vals,
longlat=FALSE, bws0, bw.seled=rep(F, length(bws0)), approach = "AIC",
bws.thresholds=rep(0.1, length(dMats)), bws.reOpts=5,verbose=F,
hatmatrix=T, predictor.centered=rep(T, length(bws0)-1), nlower = 10)
## S3 method for class 'multiscalegwr'
print(x, ...)
66 gwr.multiscale

Arguments
formula Regression model formula of a formula object
data a Spatial*DataFrame, i.e. SpatialPointsDataFrame or SpatialPolygonsDataFrame
as defined in package sp
kernel function chosen as follows:
gaussian: wgt = exp(-.5*(vdist/bw)^2);
exponential: wgt = exp(-vdist/bw);
bisquare: wgt = (1-(vdist/bw)^2)^2 if vdist < bw, wgt=0 otherwise;
tricube: wgt = (1-(vdist/bw)^3)^3 if vdist < bw, wgt=0 otherwise;
boxcar: wgt=1 if dist < bw, wgt=0 otherwise
adaptive if TRUE calculate an adaptive kernel where the bandwidth (bw) corresponds to
the number of nearest neighbours (i.e. adaptive distance); default is FALSE,
where a fixed kernel is found (bandwidth is a fixed distance)
criterion criterion for determining the convergence of the back-fitting procedure, could be
"CVR" or "dCVR", which corespond to the changing value of RSS (CVR) and
the differential version (dCVR), respectively; and "dCVR" is used as default.
max.iterations maximum number of iterations in the back-fitting procedure
threshold threshold value to terminate the back-fitting iterations
dMats a list of distance matrices used for estimating each specific parameter
p.vals a collection of positive numbers used as the power of the Minkowski distance
theta.vals a collection of values used as angles in radians to rotate the coordinate system
longlat if TRUE, great circle distances will be calculated
bws0 a vector of initializing bandwidths for the back-fitting procedure, of which the
length should equal to the number of paramters if specified
bw.seled a vector of boolean variables to determine whether the corresponding bandwidth
should be re-selected or not: if TRUE, the corresponding bandwiths for the
specific parameters are supposed to be given in bws0; otherwise, the bandwidths
for the specific parameters will be selected within the back-fitting iterations.
approach specified by CV for cross-validation approach or by AIC corrected (AICc) ap-
proach
bws.thresholds threshold values to define whether the bandwidth for a specific parameter has
converged or not
bws.reOpts the number times of continually optimizing each parameter-specific bandwidth
even though it meets the criterion of convergence, for avoiding sub-optimal
choice due to illusion of convergence;
verbose if TRUE and bandwidth selection is undertaken, the bandwidth searches are
reported
predictor.centered
a logical vector of length equalling to the number of predictors, and note inter-
cept is not included; if the element is TRUE, the corresponding predictor will be
centered.
gwr.multiscale 67

hatmatrix if TRUE the hatmatrix for the whole model will be calculated, and AICc, adjusted-
R2 values will be returned accordingly.
nlower the minmum number of nearest neighbours if an adaptive kernel is used
x an object of class “multiscalegwr”, returned by the function gwr.multiscale
... arguments passed through (unused)

Value

A list of class “psdmgwr”:

SDF a SpatialPointsDataFrame (may be gridded) or SpatialPolygonsDataFrame ob-

ject (see package “sp”) integrated with data locations,coefficient estimates from
the PSDM GWR model,predicted y values,residuals, coefficient standard errors
and t-values in its "data" slot.
GW.arguments a list class object including the model fitting parameters for generating the report
file
GW.diagnostic a list class object including the diagnostic information of the model fitting
lm an object of class inheriting from “lm”, see lm.
bws.vars bandwidths used for all the parameters within the back-fitting procedure
timings starting and ending time.
this.call the function call used.

Note

This function implements multiscale GWR to detect variations in regression relationships across
different spatial scales. This function can not only find a different bandwidth for each relationship,
but also (and simultaneously), find a different distance metric for each relationship (i.e. Parameter-
Specific Distance Metric GWR, i.e. PSDM GWR). Note that multiscale GWR (MGWR) has also
been referred to as flexible bandwidth GWR (FBGWR) and conditional GWR (CGWR) in the lit-
erature. All are one and the same model, but where PSDM-GWR additionally provides a different
distance metric option for each relationship. An MGWR model is calibrated if no “dMats” and
“p.vals” are specified; a mixed GWR model will be calibrated if an infinite bandwidth and an-
other regular bandwidth are used for estimating the global and local parameters (again when no
“dMats” and “p.vals” are specified). In other words, the gwr.multiscale function is specified with
Euclidean distances in both cases. Note that the results from this function for a mixed GWR model
and gwr.mixed might be different, as a back-fitting algorithm is used in gwr.multiscale, while an
approximating algorithm is applied in gwr.mixed. The gwr.mixed function performs better in com-
putational efficiency, but poorer in prediction accuracy.

Author(s)

Binbin Lu <[email protected]>
68 gwr.multiscale

References

Yang, W. (2014). An Extension of Geographically Weighted Regression with Flexible Bandwidths.

St Andrews, St Andrews, UK.
Lu, B., Harris, P., Charlton, M., & Brunsdon, C. (2015). Calibrating a Geographically Weighted
Regression Model with Parameter-specific Distance Metrics. Procedia Environmental Sciences, 26,
109-114.
Lu, B., Brunsdon, C., Charlton, M., & Harris, P. (2017). Geographically weighted regression with
parameter-specific distance metrics. International Journal of Geographical Information Science, 31,
982-998.
Fotheringham, A. S., Yang, W. & Kang, W. (2017). Multiscale Geographically Weighted Regres-
sion (MGWR). Annals of the American Association of Geographers, 107, 1247-1265.
Yu, H., A. S. Fotheringham, Z. Li, T. Oshan, W. Kang & L. J. Wolf. 2019. Inference in multiscale
geographically weighted regression. Geographical Analysis(In press).
Leong, Y.Y., & Yue, J.C. (2017). A modification to geographically weighted regression. Interna-
tional Journal of Health Geographics, 16 (1), 11.
Lu, B., Yang, W. Ge, Y. & Harris, P. (2018). Improvements to the calibration of a geographically
weighted regression with parameter-specific distance metrics and bandwidths. Forthcoming Com-
puters, Environment and Urban Systems.
Wolf, L.J, Oshan, T.M, Fotheringham, A.S. (2018). Single and multiscale models of process spatial
heterogeneity. Geographical Analysis, 50(3): 223-246.
Murakami, D., Lu, B., Harris, P., Brunsdon, C., Charlton, M., Nakaya, T., & Griffith, D. (2019)
The importance of scale in spatially varying coefficient modelling. Forthcoming Annals of the
Association of American Geographers.

Examples
data(LondonHP)
EUDM <- gw.dist(coordinates(londonhp))
#No bandwidth is selected, and bws0 values are used
## Not run:
###Similar as the basic GWR
res1<-gwr.multiscale(PURCHASE~FLOORSZ+PROF, data=londonhp, criterion="dCVR",kernel="gaussian",
adaptive=T, bws0=c(100, 100, 100),bw.seled=rep(T, 3), dMats=list(EUDM,EUDM,EUDM))
#FBGWR
res2<-gwr.multiscale(PURCHASE~FLOORSZ+PROF, data=londonhp, criterion="dCVR",kernel="gaussian",
adaptive=T, bws0=c(100, 100, 100), dMats=list(EUDM,EUDM,EUDM))
#Mixed GWR
res3<-gwr.multiscale(PURCHASE~FLOORSZ+PROF, data=londonhp, bws0=c(Inf, 100, 100, Inf),
bw.seled=rep(T, 3),kernel="gaussian", dMats=list(EUDM,EUDM,EUDM))
#PSDM GWR
res4<- gwr.multiscale(PURCHASE~FLOORSZ+PROF, data=londonhp, kernel="gaussian", p.vals=c(1,2,3))

## End(Not run)
gwr.predict 69

gwr.predict GWR used as a spatial predictor

Description
This function implements basic GWR as a spatial predictor. The GWR prediction function is able
to do leave-out-one predictions (when the observation locations are used for prediction) and predic-
tions at a set-aside data set (when unobserved locations are used for prediction).

Usage
gwr.predict(formula, data, predictdata, bw, kernel="bisquare",adaptive=FALSE, p=2,
theta=0, longlat=F,dMat1, dMat2)
## S3 method for class 'gwrm.pred'
print(x, ...)

Arguments
formula Regression model formula of a formula object
data a Spatial*DataFrame, i.e. SpatialPointsDataFrame or SpatialPolygonsDataFrame
as defined in package sp
predictdata a Spatial*DataFrame object to provide prediction locations, i.e. SpatialPoints-
DataFrame or SpatialPolygonsDataFrame as defined in package sp
bw bandwidth used in the weighting function, possibly calculated by bw.gwr;fixed
(distance) or adaptive bandwidth(number of nearest neighbours)
kernel function chosen as follows:
gaussian: wgt = exp(-.5*(vdist/bw)^2);
exponential: wgt = exp(-vdist/bw);
bisquare: wgt = (1-(vdist/bw)^2)^2 if vdist < bw, wgt=0 otherwise;
tricube: wgt = (1-(vdist/bw)^3)^3 if vdist < bw, wgt=0 otherwise;
boxcar: wgt=1 if dist < bw, wgt=0 otherwise
adaptive if TRUE calculate an adaptive kernel where the bandwidth (bw) corresponds to
the number of nearest neighbours (i.e. adaptive distance); default is FALSE,
where a fixed kernel is found (bandwidth is a fixed distance)
p the power of the Minkowski distance, default is 2, i.e. the Euclidean distance
theta an angle in radians to rotate the coordinate system, default is 0
longlat if TRUE, great circle distances will be calculated
dMat1 a pre-specified distance matrix between data points and prediction locations; if
not given, it will be calculated by the given parameters
dMat2 a pre-specified sysmetric distance matrix between data points; if not given, it
will be calculated by the given parameters
x an object of class “gwrm.pred”, returned by the function gwr.predict
... arguments passed through (unused)
70 gwr.predict

Value
A list of class “gwrm.pred”:

GW.arguments a list of geographically weighted arguments

SDF a SpatialPointsDataFrame (may be gridded) or SpatialPolygonsDataFrame ob-
ject (see package "sp") with GWR coefficients, predictions and prediction vari-
ances in its "data" slot.
this.call the function call used.

Author(s)
Binbin Lu <[email protected]>

References
Harris P, Fotheringham AS, Crespo R, Charlton M (2010) The use of geographically weighted
regression for spatial prediction: an evaluation of models using simulated data sets. Mathematical
Geosciences 42:657-680
Harris P, Juggins S (2011) Estimating freshwater critical load exceedance data for Great Britain
using space-varying relationship models. Mathematical Geosciences 43: 265-292
Harris P, Brunsdon C, Fotheringham AS (2011) Links, comparisons and extensions of the geo-
graphically weighted regression model when used as a spatial predictor. Stochastic Environmental
Research and Risk Assessment 25:123-138
Gollini I, Lu B, Charlton M, Brunsdon C, Harris P (2015) GWmodel: an R Package for explor-
ing Spatial Heterogeneity using Geographically Weighted Models. Journal of Statistical Software,
63(17):1-50

Examples
## Not run:
data(LondonHP)
gwr.pred<-gwr.predict(PURCHASE~FLOORSZ, data=londonhp, bw=2000,kernel = "gaussian")
gwr.pred
#########Global OLS regression results and comparison with gstat functions
if(require("gstat"))
{
mlr.g <- gstat(id = "xx1", formula = PURCHASE~FLOORSZ,data=londonhp)
mlr.g1 <- predict(mlr.g, newdata = londonhp, BLUE = TRUE)
mlr.g1
}
############
ols.pred<-gwr.predict(PURCHASE~FLOORSZ, data=londonhp, bw=100000000000000000000000)
ols.pred$SDF

## End(Not run)
gwr.robust 71

gwr.robust Robust GWR model

Description
This function implements two robust GWR models.

Usage
gwr.robust(formula, data, bw,filtered=FALSE,
kernel = "bisquare",adaptive = FALSE, p = 2,
theta = 0, longlat = F, dMat, F123.test = F,
maxiter=20,cut.filter= 3, cut1=2,cut2=3,delta=1.0e-5)

Arguments
formula Regression model formula of a formula object
data a Spatial*DataFrame, i.e. SpatialPointsDataFrame or SpatialPolygonsDataFrame
as defined in package sp
bw bandwidth used in the weighting function, possibly calculated by bw.gwr;fixed
(distance) or adaptive bandwidth(number of nearest neighbours)
filtered default FALSE, the automatic approach is used, if TRUE the filtered data ap-
proach is employed, as that described in Fotheringham et al. (2002 p.73-80)
kernel function chosen as follows:
gaussian: wgt = exp(-.5*(vdist/bw)^2);
exponential: wgt = exp(-vdist/bw);
bisquare: wgt = (1-(vdist/bw)^2)^2 if vdist < bw, wgt=0 otherwise;
tricube: wgt = (1-(vdist/bw)^3)^3 if vdist < bw, wgt=0 otherwise;
boxcar: wgt=1 if dist < bw, wgt=0 otherwise
adaptive if TRUE calculate an adaptive kernel where the bandwidth (bw) corresponds to
the number of nearest neighbours (i.e. adaptive distance); default is FALSE,
where a fixed kernel is found (bandwidth is a fixed distance)
p the power of the Minkowski distance, default is 2, i.e. the Euclidean distance
theta an angle in radians to rotate the coordinate system, default is 0
longlat if TRUE, great circle distances will be calculated
dMat a pre-specified distance matrix, it can be calculated by the function gw.dist
F123.test default FALSE, otherwise calculate F-test results (Leung et al. 2000)
maxiter default 20, maximum number of iterations for the automatic approach
cut.filter If filtered is TRUE, it will be used as the residual cutoff for filtering data; default
cutoff is 3
cut1 default 2, first cutoff for the residual weighting function. wr(e)=1 if |e|<=cut1*sigma
72 gwr.robust

cut2 default 3, second cutoff for the residual weighting function. wr(e)=(1-(|e|-2)^2)^2
if cut1*sigma<|e|<cut2*sigma, and wr(e)=0 if |e|>=cut2*sigma; cut 1 and cut2
refer to the automatic approach
delta default 1.0e-5, tolerance of the iterative algorithm

Value
A list of class “gwrm”:

GW.arguments a list class object including the model fitting parameters for generating the report
file
GW.diagnostic a list class object including the diagnostic information of the model fitting
lm an object of class inheriting from “lm”, see lm.
SDF a SpatialPointsDataFrame (may be gridded) or SpatialPolygonsDataFrame ob-
ject (see package “sp”) integrated with fit.points,GWR coefficient estimates, y
value,predicted values, coefficient standard errors and t-values in its "data" slot.
Notably, E_weigts will be also included in the output SDF which represents the
residual weighting when automatic approach is used; When the filtered approach
is used, E_weight is a vector consisted of 0 and 1, where 0 means outlier to be
excluded from calibration.
timings starting and ending time.
this.call the function call used.
Ftest.res results of Leung’s F tests when F123.test is TRUE.

Author(s)
Binbin Lu <[email protected]>

References
Fotheringham S, Brunsdon, C, and Charlton, M (2002), Geographically Weighted Regression: The
Analysis of Spatially Varying Relationships, Chichester: Wiley.
Harris P, Fotheringham AS, Juggins S (2010) Robust geographically weighed regression: a tech-
nique for quantifying spatial relationships between freshwater acidification critical loads and catch-
ment attributes. Annals of the Association of American Geographers 100(2): 286-306

Examples
## Not run:
data(DubVoter)
bw.a <- bw.gwr(GenEl2004~DiffAdd+LARent+SC1+Unempl+LowEduc+Age18_24
+Age25_44+Age45_64,
data=Dub.voter,approach="AICc",kernel="bisquare",adaptive=TRUE)
bw.a
gwr.res <- gwr.basic(GenEl2004~DiffAdd+LARent+SC1+Unempl+LowEduc+Age18_24
+Age25_44+Age45_64,
data=Dub.voter,bw=bw.a,kernel="bisquare",adaptive=TRUE,F123.test=TRUE)
print(gwr.res)
gwr.scalable 73

# Map of the estimated coefficients for LowEduc

names(gwr.res$SDF)
if(require("RColorBrewer"))
{
mypalette<-brewer.pal(6,"Spectral")
X11(width=10,height=12)
spplot(gwr.res$SDF,"LowEduc",key.space = "right",
col.regions=mypalette,at=c(-8,-6,-4,-2,0,2,4),
main="Basic GW regression coefficient estimates for LowEduc")
}
# Robust GW regression and map of the estimated coefficients for LowEduc
rgwr.res <- gwr.robust(GenEl2004~DiffAdd+LARent+SC1+Unempl+LowEduc+Age18_24
+Age25_44+Age45_64, data=Dub.voter,bw=bw.a,kernel="bisquare",
adaptive=TRUE,F123.test=TRUE)
print(rgwr.res)
if(require("RColorBrewer"))
{
X11(width=10,height=12)
spplot(rgwr.res$SDF, "LowEduc", key.space = "right",
col.regions=mypalette,at=c(-8,-6,-4,-2,0,2,4),
main="Robust GW regression coefficient estimates for LowEduc")
}

## End(Not run)

gwr.scalable Scalable GWR

Description
This function implements Scalable GWR for large dataset

Usage
gwr.scalable(formula, data, bw.adapt=100, kernel = "gaussian", polynomial = 4,
p = 2, theta = 0, longlat = F, dMat)
## S3 method for class 'scgwrm'
print(x, ...)

kernel Kernel function to calculate the spatial weights, but note only two continuous
functions available:
gaussian: wgt = exp(-.5*(vdist/bw)^2);
exponential: wgt = exp(-vdist/bw);
polynomial Degree of the polyunomial to approximate the kernel function, and default is 4.
p the power of the Minkowski distance, default is 2, i.e. the Euclidean distance
theta an angle in radians to rotate the coordinate system, default is 0
longlat if TRUE, great circle distances will be calculated
dMat a pre-specified distance matrix, it can be calculated by the function gw.dist
x an object of class “scgwrm”, returned by the function gwr.scalable
... arguments passed through (unused)

Value
A list of class “scgwrm”:

Author(s)
Binbin Lu <[email protected]>

References
Murakami, D., N. Tsutsumida, T. Yoshida, T. Nakaya & B. Lu (2019) Scalable GWR: A linear-time
algorithm for large-scale geographically weighted regression with polynomial kernels. arXiv:1905.00266.

Examples
## Not run:
require(spData)
data(boston)
boston <- boston.c
coordinates(boston) <- ~ LON + LAT
res <- gwr.scalable(formula = MEDV ~ CRIM + ZN + INDUS + CHAS + AGE, data = boston, bw.adapt = 100)
res

## End(Not run)
gwr.t.adjust 75

gwr.t.adjust Adjust p-values for multiple hypothesis tests in basic GWR

Description
Given a set of p-values from the pseudo t-tests of basic GWR outputs, this function returns ad-
justed p-values using: (a) Bonferroni, (b) Benjamini-Hochberg, (c) Benjamini-Yekutieli and (d)
Fotheringham-Byrne procedures.

Usage
gwr.t.adjust(gwm.Obj)

Arguments
gwm.Obj an object of class “gwrm”, returned by the function gwr.basic

Author(s)
Binbin Lu <[email protected]>

References
Byrne, G., Charlton, M. and Fotheringham, S., 2009. Multiple dependent hypothesis tests in geo-
graphically weighted regression. In: Lees, B. and Laffan, S. eds. 10th International conference on
geocomputation. Sydney.

gwr.write Write the GWR results into files

Description
This function writes the calibration result of function gwr.basic to a text file and shape files

Usage
gwr.write(x,fn="GWRresults")
gwr.write.shp(x,fn="GWRresults")

Arguments
x an object of class “gwrm”, returned by the function gwr.basic
fn file name for the written results, by default the output files can be found in the
working directory, “GWRresults.txt”, “GWRresults(.shp, .shx, .dbf)”
76 gwss

Note
The projection file is missing for the writen shapefiles.
The functions “writeGWR” and “writeGWR.shp” (in the early versions of GWmodel) have been
renamed respectively as “gwr.write” and “gwr.write.shp”, while the old names are still kept valid.

Author(s)
Binbin Lu <[email protected]>

gwss Geographically weighted summary statistics (GWSS)

Description
This function calculates basic and robust GWSS. This includes geographically weighted means,
standard deviations and skew. Robust alternatives include geographically weighted medians, inter-
quartile ranges and quantile imbalances. This function also calculates basic geographically weighted
covariances together with basic and robust geographically weighted correlations.

Usage
gwss(data, summary.locat,vars,kernel="bisquare",adaptive=FALSE, bw,p=2,
theta=0, longlat=F,dMat,quantile=FALSE)
## S3 method for class 'gwss'
print(x, ...)

Arguments
data a Spatial*DataFrame, i.e. SpatialPointsDataFrame or SpatialPolygonsDataFrame
as defined in package sp
summary.locat a Spatial*DataFrame object for providing summary locations, i.e. SpatialPoints-
DataFrame or SpatialPolygonsDataFrame as defined in package sp
vars a vector of variable names to be summarized
bw bandwidth used in the weighting function
kernel function chosen as follows:
gaussian: wgt = exp(-.5*(vdist/bw)^2);
exponential: wgt = exp(-vdist/bw);
bisquare: wgt = (1-(vdist/bw)^2)^2 if vdist < bw, wgt=0 otherwise;
tricube: wgt = (1-(vdist/bw)^3)^3 if vdist < bw, wgt=0 otherwise;
boxcar: wgt=1 if dist < bw, wgt=0 otherwise
adaptive if TRUE calculate an adaptive kernel where the bandwidth (bw) corresponds to
the number of nearest neighbours (i.e. adaptive distance); default is FALSE,
where a fixed kernel is found (bandwidth is a fixed distance)
p the power of the Minkowski distance, default is 2, i.e. the Euclidean distance
gwss 77

theta an angle in radians to rotate the coordinate system, default is 0

longlat if TRUE, great circle distances will be calculated
dMat a pre-specified distance matrix, it can be calculated by the function gw.dist
quantile if TRUE, median, interquartile range, quantile imbalance will be calculated
x an object of class “gwss”, returned by the function gwss
... arguments passed through (unused)

Value
A list of class “lss”:

SDF a SpatialPointsDataFrame (may be gridded) or SpatialPolygonsDataFrame ob-

ject (see package “sp”) with local means,local standard deviations,local vari-
ance, local skew,local coefficients of variation, local covariances, local cor-
relations (Pearson’s), local correlations (Spearman’s), local medians, local in-
terquartile ranges, local quantile imbalances and coordinates.
... other information for reporting

Author(s)
Binbin Lu <[email protected]>

References
Fotheringham S, Brunsdon, C, and Charlton, M (2002), Geographically Weighted Regression: The
Analysis of Spatially Varying Relationships, Chichester: Wiley.
Brunsdon C, Fotheringham AS, Charlton ME (2002) Geographically weighted summary statistics -
a framework for localised exploratory data analysis. Computers, Environment and Urban Systems
26:501-524
Harris P, Clarke A, Juggins S, Brunsdon C, Charlton M (2014) Geographically weighted meth-
ods and their use in network re-designs for environmental monitoring. Stochastic Environmental
Research and Risk Assessment 28: 1869-1887

Examples
## Not run:
data(EWHP)
data(EWOutline)
head(ewhp)
houses.spdf <- SpatialPointsDataFrame(ewhp[, 1:2], ewhp)
localstats1 <- gwss(houses.spdf, vars = c("PurPrice", "FlrArea"), bw = 50000)
head(data.frame(localstats1$SDF))
localstats1
##A function for mapping data
if(require("RColorBrewer"))
{
quick.map <- function(spdf,var,legend.title,main.title)
{
78 gwss.montecarlo

x <- spdf@data[,var]
cut.vals <- pretty(x)
x.cut <- cut(x,cut.vals)
cut.levels <- levels(x.cut)
cut.band <- match(x.cut,cut.levels)
colors <- brewer.pal(length(cut.levels), "YlOrRd")
colors <- rev(colors)
par(mar=c(1,1,1,1))
plot(ewoutline,col="olivedrab",bg="lightblue1")
title(main.title)
plot(spdf,add=TRUE,col=colors[cut.band],pch=16)
legend("topleft",cut.levels,col=colors,pch=16,bty="n",title=legend.title)
}
quick.map(localstats1$SDF, "PurPrice_LM", "1000's Uk Pounds",
"Geographically Weighted Mean")
par(mfrow = c(1, 2))
quick.map(localstats1$SDF, "PurPrice_LSKe", "Skewness Level", "Local Skewness")
quick.map(localstats1$SDF, "PurPrice_LSD", "1000's Pounds", "Local Standard Deviation")
#Exploring Non-Stationarity of Relationships
quick.map(localstats1$SDF, "Corr_PurPrice.FlrArea", expression(rho),
"Geographically Weighted Pearson Correlation")
#Robust, Quantile Based Local Summary Statistics
localstats2 <- gwss(houses.spdf, vars = c("PurPrice", "FlrArea"),
bw = 50000, quantile = TRUE)
quick.map(localstats2$SDF, "PurPrice_Median", "1000 UK Pounds",
"Geographically Weighted Median House Price")
}

## End(Not run)

gwss.montecarlo Monte Carlo (randomisation) test for gwss

Description
This function implements Monte Carlo (randomisation) tests for the GW summary statistics found
in gwss.

Usage
gwss.montecarlo(data, vars, kernel = "bisquare",
adaptive = FALSE, bw, p = 2, theta = 0, longlat = F,
dMat, quantile=FALSE,nsim=99)

Arguments
data a Spatial*DataFrame, i.e. SpatialPointsDataFrame or SpatialPolygonsDataFrame
as defined in package sp
vars a vector of variable names to be summarized
gwss.montecarlo 79

bw bandwidth used in the weighting function

kernel function chosen as follows:
gaussian: wgt = exp(-.5*(vdist/bw)^2);
exponential: wgt = exp(-vdist/bw);
bisquare: wgt = (1-(vdist/bw)^2)^2 if vdist < bw, wgt=0 otherwise;
tricube: wgt = (1-(vdist/bw)^3)^3 if vdist < bw, wgt=0 otherwise;
boxcar: wgt=1 if dist < bw, wgt=0 otherwise
adaptive if TRUE calulate the adaptive kernel, and bw correspond to the number of near-
est neighbours, default is FALSE.
p the power of the Minkowski distance, default is 2, i.e. the Euclidean distance
theta an angle in radians to rotate the coordinate system, default is 0
longlat if TRUE, great circle distances will be calculated
dMat a pre-specified distance matrix, it can be calculated by the function gw.dist
quantile if TRUE, median, interquartile range, quantile imbalance will be calculated
nsim default 99, the number of randomisations

Value

test probability of the test statistics of the GW summary statistics; if p<0.025 or if

p>0.975 then the true local summary statistics can be said to be significantly
different (at the 0.95 level) to such a local summary statistics found by chance.

Note

The function “montecarlo.gwss” (in the early versions of GWmodel) has been renamed as “gwss.montecarlo”,
while the old name is still kept valid.

Author(s)

Binbin Lu <[email protected]>

References

Fotheringham S, Brunsdon, C, and Charlton, M (2002), Geographically Weighted Regression: The

Analysis of Spatially Varying Relationships, Chichester: Wiley.
Brunsdon C, Fotheringham AS, Charlton ME (2002) Geographically weighted summary statistics -
a framework for localised exploratory data analysis. Computers, Environment and Urban Systems
26:501-524
Harris P, Brunsdon C (2010) Exploring spatial variation and spatial relationships in a freshwater
acidification critical load data set for Great Britain using geographically weighted summary statis-
tics. Computers & Geosciences 36:54-70
80 LondonHP

Examples
## Not run:
data(LondonHP)
DM<-gw.dist(dp.locat=coordinates(londonhp))
test.lss<-gwss.montecarlo(data=londonhp, vars=c("PURCHASE","FLOORSZ"), bw=5000,
kernel ="gaussian", dMat=DM,nsim=99)
test.lss

## End(Not run)

LondonBorough London boroughs data

Description
Outline (SpatialPolygonsDataFrame) of London boroughs for the LondonHP data.

Usage
data(LondonBorough)

Author(s)
Binbin Lu <[email protected]>

LondonHP London house price data set (SpatialPointsDataFrame)

Description
A house price data set with 18 hedonic variables for London in 2001.

Usage
data(LondonHP)

Format
A SpatialPointsDataFrame object (proj4string set to "+init=epsg:27700 +datum=OSGB36").
The "data" slot is a data frame with 372 observations on the following 21 variables.
X a numeric vector, X coordinate
Y a numeric vector, Y coordinate
PURCHASE a numeric vector, the purchase price of the property
FLOORSZ a numeric vector, floor area of the property in square metres
LondonHP 81

TYPEDETCH a numeric vector, 1 if the property is detached (i.e. it is a stand-alone house), 0

otherwise
TPSEMIDTCH a numeric vector, 1 if the property is semi detached, 0 otherwise
TYPETRRD a numeric vector, 1 if the property is in a terrace of similar houses (commonly re-
ferred to as a ’row house’ in the USA), 0 otherwise
TYPEBNGLW a numeric vector, if the property is a bungalow (i.e. it has only one floor), 0
otherwise
TYPEFLAT a numeric vector, if the property is a flat (or ’apartment’ in the USA), 0 otherwise
BLDPWW1 a numeric vector, 1 if the property was built prior to 1914, 0 otherwise
BLDPOSTW a numeric vector, 1 if the property was built between 1940 and 1959, 0 otherwise
BLD60S a numeric vector, 1 if the property was built between 1960 and 1969, 0 otherwise
BLD70S a numeric vector, 1 if the property was built between 1970 and 1979, 0 otherwise
BLD80S a numeric vector, 1 if the property was built between 1980 and 1989, 0 otherwise
BLD90S a numeric vector, 1 if the property was built between 1990 and 2000, 0 otherwise
BATH2 a numeric vector, 1 if the property has more than 2 bathrooms, 0 otherwise
GARAGE a numeric vector,1 if the house has a garage, 0 otherwise
CENTHEAT a numeric vector, 1 if the house has central heating, 0 otherwise
BEDS2 a numeric vector, 1 if the property has more than 2 bedrooms, 0 otherwise
UNEMPLOY a numeric vector, the rate of unemployment in the census ward in which the house
is located
PROF a numeric vector, the proportion of the workforce in professional or managerial occupations
in the census ward in which the house is located

Author(s)
Binbin Lu <[email protected]>

References
Fotheringham, A.S., Brunsdon, C., and Charlton, M.E. (2002), Geographically Weighted Regres-
sion: The Analysis of Spatially Varying Relationships, Chichester: Wiley.
Lu, B, Charlton, M, Harris, P, Fotheringham, AS (2014) Geographically weighted regression with
a non-Euclidean distance metric: a case study using hedonic house price data. International Journal
of Geographical Information Science 28(4): 660-681

Examples
data(LondonHP)
data(LondonBorough)
ls()
plot(londonborough)
plot(londonhp, add=TRUE)
82 USelect

USelect Results of the 2004 US presidential election at the county level (Spa-
tialPolygonsDataFrame)

Description
Results of the 2004 US presidential election at the county level, together with five socio-economic
(census) variables. This data can be used with GW Discriminant Analysis.

Usage
data(USelect)

Format
A SpatialPolygonsDataFrame with 3111 electoral divisions on the following 6 variables.
winner Categorical variable with three classes: i) Bush, ii) Kerry and iii) Borderline (supporting
ratio for a candidate ranges from 0.45 to 0.55)
unemploy percentage unemployed
pctcoled percentage of adults over 25 with 4 or more years of college education
PEROVER65 percentage of persons over the age of 65
pcturban percentage urban
WHITE percentage white

References
Robinson, A. C. (2013). Geovisualization of the 2004 Presidential Election. In: NATIONAL IN-
STITUTES OF HEALTH, P. S. U. (ed.). Penn State: http://www.personal.psu.edu/users/a/
c/acr181/election.html.
Foley, P. & Demsar, U. (2012). Using geovisual analytics to compare the performance of geograph-
ically weighted discriminant analysis versus its global counterpart, linear discriminant analysis.
International Journal of Geographical Information Science, 27, 633-661.

Examples

data(USelect)
ls()
Index

∗Topic GTWR gwr.t.adjust, 75

bw.gtwr, 5 gwr.write, 75
gtwr, 21 ∗Topic GWSS
∗Topic GW tools bw.gwss.average, 12
gw.dist, 23 gwss, 76
gw.pcplot, 24 gwss.montecarlo, 78
gw.weight, 26 ∗Topic Heteroskedastic GWR
∗Topic GWDA gwr.hetero, 48
bw.gwda, 7 ∗Topic Scalable GWR
gwda, 27 gwr.scalable, 73
∗Topic GWPCA ∗Topic data
bw.gwpca, 8 DubVoter, 13
gwpca, 28 EWHP, 14
gwpca.check.components, 32 EWOutline, 15
gwpca.cv, 32 Georgia, 15
gwpca.cv.contrib, 33 GeorgiaCounties, 16
gwpca.glyph.plot, 34 LondonBorough, 80
gwpca.montecarlo.1, 35 LondonHP, 80
gwpca.montecarlo.2, 37 USelect, 82
∗Topic GWR-LCR ∗Topic generalised GWR
bw.gwr.lcr, 11 bw.ggwr, 4
ggwr.basic, 17
gwr.lcr, 49
ggwr.cv, 19
gwr.lcr.cv, 51
ggwr.cv.contrib, 20
gwr.lcr.cv.contrib, 53
∗Topic multiscale GWR
∗Topic GWR
gwr.mixed, 58
bw.gwr, 9
gwr.multiscale, 65
gwr.basic, 38
∗Topic p-values adjustment
gwr.bootstrap, 41
gwr.t.adjust, 75
gwr.collin.diagno, 44
∗Topic package
gwr.cv, 46 GWmodel-package, 3
gwr.cv.contrib, 47 ∗Topic robust GWR
gwr.mink.approach, 54 gwr.robust, 71
gwr.mink.matrixview, 56
gwr.mink.pval, 56 AICc (gwr.model.selection), 60
gwr.model.selection, 60 AICc1 (gwr.scalable), 73
gwr.model.sort, 62 AICc_rss (gwr.model.selection), 60
gwr.model.view, 63
gwr.montecarlo, 64 bias.bs (gwr.bootstrap), 41
gwr.predict, 69 bisq_wt_mat (gw.weight), 26

83
84 INDEX

bisq_wt_vec (gw.weight), 26 generate.lm.data (gwr.bootstrap), 41

bw.ggwr, 4 Georgia, 15
bw.gtwr, 5 GeorgiaCounties, 16
bw.gwda, 7 get.before.ti (gtwr), 21
bw.gwpca, 8, 28, 29, 35 get.ts (gtwr), 21
bw.gwr, 9, 22, 26, 39, 48, 54, 57, 59, 60, 64, get.uloat (gtwr), 21
69, 71 ggwr.aic (bw.ggwr), 4
bw.gwr.lcr, 11 ggwr.basic, 17
bw.gwr1 (gwr.mink.approach), 54 ggwr.cv, 19
bw.gwr3 (gwr.bootstrap), 41 ggwr.cv.contrib, 20
bw.gwss.average, 12 glm, 18
glyph.plot (gwpca.glyph.plot), 34
cd_dist_mat (gw.dist), 23 gold (bw.gwr), 9
cd_dist_smat (gw.dist), 23 grouping.xy (gwda), 27
cd_dist_vec (gw.dist), 23 gtwr, 21
check.components gtwr.aic (bw.gtwr), 5
(gwpca.check.components), 32 gtwr.cv (bw.gtwr), 5
ci.bs (gwr.bootstrap), 41 gw.average.cv (bw.gwss.average), 12
Ci_mat (gwr.basic), 38 gw.dist, 5, 7, 9–11, 13, 18, 20, 21, 23, 25, 28,
confusion.matrix (gwda), 27 29, 33, 34, 36, 38, 39, 42, 45–48, 50,
coordinate_rotate (gw.dist), 23 52, 54, 59, 61, 64, 71, 74, 77, 79
gw.fitted (gwr.model.selection), 60
dist, 24
gw.mean.cv (bw.gwss.average), 12
Dub.voter (DubVoter), 13
gw.median.cv (bw.gwss.average), 12
DubVoter, 13
gw.pcplot, 24
gw.reg1 (gwr.predict), 69
ehat (gwr.model.selection), 60
gw.weight, 26
eu_dist_mat (gw.dist), 23
gw_reg (gwr.basic), 38
eu_dist_smat (gw.dist), 23
gwda, 27
eu_dist_vec (gw.dist), 23
EWHP, 14, 15 GWmodel (GWmodel-package), 3
ewhp (EWHP), 14 GWmodel-package, 3
EWOutline, 15 gwpca, 28, 29, 32, 35
ewoutline (EWOutline), 15 gwpca.check.components, 31
exp_wt_mat (gw.weight), 26 gwpca.cv, 32
exp_wt_vec (gw.weight), 26 gwpca.cv.contrib, 33
extract.mat (gwr.model.selection), 60 gwpca.glyph.plot, 32, 34
gwpca.montecarlo.1, 35, 36
F1234.test (gwr.basic), 38 gwpca.montecarlo.2, 36, 37
formula, 4, 5, 7, 10, 11, 17, 22, 27, 39, 41, 42, gwr.aic (bw.gwr), 9
45, 48, 49, 54, 57, 59, 64, 66, 69, 71, gwr.aic1 (gwr.mink.approach), 54
73 gwr.backfit (gwr.multiscale), 65
gwr.basic, 38, 39, 75
gauss_wt_mat (gw.weight), 26 gwr.binomial (ggwr.basic), 17
gauss_wt_vec (gw.weight), 26 gwr.bootstrap, 41, 42
Gedu.counties (GeorgiaCounties), 16 gwr.collin.diagno, 44
Gedu.df (Georgia), 15 gwr.cv, 46
Generate.formula (gwr.model.selection), gwr.cv.contrib, 47
60 gwr.cv1 (gwr.mink.approach), 54
INDEX 85

gwr.fitted (ggwr.basic), 17 mk_dist_smat (gw.dist), 23

gwr.generalised, 18 mk_dist_vec (gw.dist), 23
gwr.generalised (ggwr.basic), 17 model.selection.gwr
gwr.hetero, 48 (gwr.model.selection), 60
gwr.lcr, 11, 49, 50 model.sort.gwr (gwr.model.sort), 62
gwr.lcr.cv, 51 model.view.gwr (gwr.model.view), 63
gwr.lcr.cv.contrib, 53 montecarlo.gwpca.1
gwr.mink.approach, 54, 56 (gwpca.montecarlo.1), 35
gwr.mink.matrixview, 56 montecarlo.gwpca.2
gwr.mink.pval, 56 (gwpca.montecarlo.2), 37
gwr.mixed, 58, 67 montecarlo.gwr (gwr.montecarlo), 64
gwr.model.selection, 60, 62, 63 montecarlo.gwss (gwss.montecarlo), 78
gwr.model.sort, 62, 62, 63
gwr.model.view, 62, 63 par, 25
gwr.montecarlo, 64 parametric.bs (gwr.bootstrap), 41
gwr.multiscale, 65, 67 plot.mcsims (gwpca.montecarlo.1), 35
gwr.poisson (ggwr.basic), 17 plot.pvlas (gwr.mink.pval), 56
gwr.predict, 69, 69 POSIXlt, 5, 22
gwr.q (gwr.mixed), 58 princomp, 29
gwr.q2 (gwr.multiscale), 65 print.ggwrm (ggwr.basic), 17
gwr.robust, 71 print.gtwrm (gtwr), 21
gwr.scalable, 73, 74 print.gwda (gwda), 27
gwr.t.adjust, 75 print.gwpca (gwpca), 28
gwr.write, 75 print.gwrbsm (gwr.bootstrap), 41
gwr_diag (gwr.basic), 38 print.gwrlcr (gwr.lcr), 49
gwr_diag1 (gwr.scalable), 73 print.gwrm (gwr.basic), 38
gwrt.err (gwr.bootstrap), 41 print.gwrm.pred (gwr.predict), 69
gwrt.lag (gwr.bootstrap), 41 print.gwss (gwss), 76
gwrt.mlr (gwr.bootstrap), 41 print.mgwr (gwr.mixed), 58
gwrt.sma (gwr.bootstrap), 41 print.multiscalegwr (gwr.multiscale), 65
gwrtvar (gwr.bootstrap), 41 print.scgwrm (gwr.scalable), 73
gwss, 76, 77, 78 pval.bs (gwr.bootstrap), 41
gwss.montecarlo, 78
ridge.lm (gwr.lcr), 49
list, 18, 66, 72 robustSvd (gwpca), 28
lm, 22, 40, 67, 72, 74 rss (gwr.model.selection), 60
local.corr (gwss), 76 rwpca (gwpca), 28
LondonBorough, 80
londonborough (LondonBorough), 80 scgwr_loocv (gwr.scalable), 73
LondonHP, 80, 80 scgwr_pre (gwr.scalable), 73
londonhp (LondonHP), 80 scgwr_reg (gwr.scalable), 73
sdist.mat (gtwr), 21
md_dist_mat (gw.dist), 23 se.bs (gwr.bootstrap), 41
md_dist_smat (gw.dist), 23 splitx (gwda), 27
md_dist_vec (gw.dist), 23 st.dist (gtwr), 21
mink.approach (gwr.mink.approach), 54
mink.matrixview (gwr.mink.matrixview), tdist.mat (gtwr), 21
56 ti.dist (gtwr), 21
mk_dist_mat (gw.dist), 23 ti.distm (gtwr), 21
86 INDEX

ti.distv (gtwr), 21
tri_wt_mat (gw.weight), 26
tri_wt_vec (gw.weight), 26

USelect, 82
USelect2004 (USelect), 82

vector, 66

wlda (gwda), 27
wlda.cr (bw.gwda), 7
wmean (gwda), 27
wpca (gwpca), 28
wprior (gwda), 27
wqda (gwda), 27
wqda.cr (bw.gwda), 7
writeGWR (gwr.write), 75
wt.median (gwpca), 28
wvarcov (gwda), 27

Intrusion Detection Honeypots
From Everand
Intrusion Detection Honeypots
Chris Sanders
3/5 (2)
Solution Manual: Linear Algebra and Its Applications, 2ed
86% (7)
Solution Manual: Linear Algebra and Its Applications, 2ed
61 pages
Oneness
From Everand
Oneness
Carmeline Pusateri
No ratings yet
A Man's Guide: Navigating the menstrual mood swing
From Everand
A Man's Guide: Navigating the menstrual mood swing
Meghan Kurts-Forrester
No ratings yet
Advanced Linear Algebra PDF
100% (13)
Advanced Linear Algebra PDF
348 pages
07 Bubble Break
No ratings yet
07 Bubble Break
25 pages
GWmodel
No ratings yet
GWmodel
70 pages
Knit Soxx for Everyone: 25 Colorful Sock Patterns for the Whole Family
From Everand
Knit Soxx for Everyone: 25 Colorful Sock Patterns for the Whole Family
Kerstin Balke
4.5/5 (2)
Knots, Bends, Splices With tables of strengths of ropes, etc. and wire rigging
From Everand
Knots, Bends, Splices With tables of strengths of ropes, etc. and wire rigging
Captain Jutsum
No ratings yet
Fly Fishing Guide to the Battenkill: Complete Guide to Locations, Hatches, and History
From Everand
Fly Fishing Guide to the Battenkill: Complete Guide to Locations, Hatches, and History
Doug Lyons
No ratings yet
The Gracious Lily Affair
From Everand
The Gracious Lily Affair
Van Wyck Mason
5/5 (1)
The Attention Fix: How to Focus in a World That Wants to Distract You
From Everand
The Attention Fix: How to Focus in a World That Wants to Distract You
Anders Hansen
No ratings yet
POTS!!!: No-Frills Recipes
From Everand
POTS!!!: No-Frills Recipes
Stephen Davis
No ratings yet
Deadline Istanbul (The Elizabeth Darcy Series)
From Everand
Deadline Istanbul (The Elizabeth Darcy Series)
Peggy Hanson
5/5 (1)
Dismo
No ratings yet
Dismo
69 pages
The Essential Robert Frost: Collected Poems 1913-1923 (Warbler Classics Annotated Edition)
From Everand
The Essential Robert Frost: Collected Poems 1913-1923 (Warbler Classics Annotated Edition)
Robert Frost
No ratings yet
Osama the Gun
From Everand
Osama the Gun
Norman Spinrad
5/5 (1)
Hamlet Had an Uncle: A Comedy of Honor
From Everand
Hamlet Had an Uncle: A Comedy of Honor
James Branch Cabell
4.5/5 (7)
A Field Guide to the Wildlife of South Georgia
From Everand
A Field Guide to the Wildlife of South Georgia
Robert Burton
4/5 (1)
Keys to Better Reading
From Everand
Keys to Better Reading
Judy McFall
No ratings yet
Deadline Yemen (The Elizabeth Darcy Series)
From Everand
Deadline Yemen (The Elizabeth Darcy Series)
Peggy Hanson
5/5 (1)
Between River and Mountain
From Everand
Between River and Mountain
Sally Walker Brinkmann
No ratings yet
Bimbo Heaven: Stone Angel #7
From Everand
Bimbo Heaven: Stone Angel #7
Marvin H. Albert
No ratings yet
Life Ain't A Dress Rehearsal: Lives in Poetry
From Everand
Life Ain't A Dress Rehearsal: Lives in Poetry
Cecil D. Haas
No ratings yet
Complete Guide to Paper Clay: Mixing recipes; Building, finishing and firing; 10 practice projects
From Everand
Complete Guide to Paper Clay: Mixing recipes; Building, finishing and firing; 10 practice projects
Liliane Tardio-Brise
No ratings yet
Tigers in the Mud: The Combat Career of German Panzer Commander Otto Carius
From Everand
Tigers in the Mud: The Combat Career of German Panzer Commander Otto Carius
Otto Carius
4/5 (64)
British and Irish Wild Flowers and Plants: A Pocket Guide
From Everand
British and Irish Wild Flowers and Plants: A Pocket Guide
Rachel Hamilton
No ratings yet
The Future Is Ours: The Collected Science Fiction of Edward D. Hoch
From Everand
The Future Is Ours: The Collected Science Fiction of Edward D. Hoch
Edward D. Hoch
No ratings yet
Operation Exile
From Everand
Operation Exile
E. Hoffmann Price
3.5/5 (1)
Anne of green gables
From Everand
Anne of green gables
Lucy Maud Montgomery
No ratings yet
Great British Bike Rides: 40 Classic routes for road cyclists
From Everand
Great British Bike Rides: 40 Classic routes for road cyclists
Dave Barter
No ratings yet
The Last Smile: Stone Angel #5
From Everand
The Last Smile: Stone Angel #5
Marvin H. Albert
No ratings yet
The Blue Ribbon Country Cookbook
From Everand
The Blue Ribbon Country Cookbook
Diane Roupe
5/5 (4)
Fly Fishing Guide to the Colorado River and Tributaries: Hatches, Fly Patterns, and Guide's Advice
From Everand
Fly Fishing Guide to the Colorado River and Tributaries: Hatches, Fly Patterns, and Guide's Advice
Bob Dye
No ratings yet
Murder in the Willett Family: A Lt. Valcour Mystery
From Everand
Murder in the Willett Family: A Lt. Valcour Mystery
Rufus King
No ratings yet
Introduction To Geographically Weighted Regression (GWR) (PDFDrive)
No ratings yet
Introduction To Geographically Weighted Regression (GWR) (PDFDrive)
58 pages
The Maxim Gorky MEGAPACK®: 61 Classic Novels and Stories
From Everand
The Maxim Gorky MEGAPACK®: 61 Classic Novels and Stories
Maxim Gorky
No ratings yet
The Bjørnstjerne Bjørnson MEGAPACK ®: 159 Classic Works
From Everand
The Bjørnstjerne Bjørnson MEGAPACK ®: 159 Classic Works
Bjørnstjerne Bjørnson
No ratings yet
Gray Hat Hacking the Ethical Hacker's
From Everand
Gray Hat Hacking the Ethical Hacker's
Çağatay Şanlı
5/5 (1)
Operation Longlife
From Everand
Operation Longlife
E. Hoffmann Price
3.5/5 (3)
Aquaponics How to do Everything from Backyard to Profitable Business: from BACKYARD to PROFITABLE BUSINESS
From Everand
Aquaponics How to do Everything from Backyard to Profitable Business: from BACKYARD to PROFITABLE BUSINESS
David H Dudley
No ratings yet
The Katherine Mansfield MEGAPACK ®: 101 Classic Works
From Everand
The Katherine Mansfield MEGAPACK ®: 101 Classic Works
Katherine Mansfield
No ratings yet
Hideous Faces, Beautiful Skulls: Tales of Horror and the Bizarre
From Everand
Hideous Faces, Beautiful Skulls: Tales of Horror and the Bizarre
Mark McLaughlin
5/5 (1)
SENIOR CITIZEN LIVING: Discover How You Can Make Your Senior Years As Golden as Possible!
From Everand
SENIOR CITIZEN LIVING: Discover How You Can Make Your Senior Years As Golden as Possible!
Karllo MELLO
No ratings yet
Tswge
No ratings yet
Tswge
157 pages
A Home in Bloom: Four Enchanted Seasons with Flowers
From Everand
A Home in Bloom: Four Enchanted Seasons with Flowers
Christie Purifoy
No ratings yet
Never Walk Alone
From Everand
Never Walk Alone
Rufus King
No ratings yet
My FIlipino Cuisines and Asian Delights that I Love!
From Everand
My FIlipino Cuisines and Asian Delights that I Love!
Yolanda Ballard
No ratings yet
Trouble in Tahiti: Blood on the Hibiscus
From Everand
Trouble in Tahiti: Blood on the Hibiscus
Hayford Peirce
No ratings yet
An Accidental Jewel: Wisconsin's Turtle-Flambeau Flowage
From Everand
An Accidental Jewel: Wisconsin's Turtle-Flambeau Flowage
Michael Hittle
No ratings yet
Geographically Weighted Regression Workbook (PDFDrive)
No ratings yet
Geographically Weighted Regression Workbook (PDFDrive)
75 pages
Duenna to a Murder
From Everand
Duenna to a Murder
Rufus King
No ratings yet
Aquaponics Design Plans, Construction, Operation, and Income: Organic Food
From Everand
Aquaponics Design Plans, Construction, Operation, and Income: Organic Food
David H Dudley
No ratings yet
Content Creation Revolution with chatGPT
From Everand
Content Creation Revolution with chatGPT
Maria Cowen
No ratings yet
The Naked Eye
From Everand
The Naked Eye
Henriette Martin
No ratings yet
Kellory the Warlock
From Everand
Kellory the Warlock
Lin Carter
No ratings yet
Gnuplot 5.2
No ratings yet
Gnuplot 5.2
271 pages
Gnuplot PDF
No ratings yet
Gnuplot PDF
269 pages
Gnu Plot
No ratings yet
Gnu Plot
258 pages
Gnuplot PDF
No ratings yet
Gnuplot PDF
254 pages
GMT Docs PDF
No ratings yet
GMT Docs PDF
318 pages
Withson Gnu2
No ratings yet
Withson Gnu2
238 pages
Gnuplot 4.6: An Interactive Plotting Program
No ratings yet
Gnuplot 4.6: An Interactive Plotting Program
238 pages
JAM Mathematics Sample Paper 2019 PDF
No ratings yet
JAM Mathematics Sample Paper 2019 PDF
14 pages
Least Squares Solution and Pseudo-Inverse: Bghiggins/Ucdavis/Ech256/Jan - 2012
No ratings yet
Least Squares Solution and Pseudo-Inverse: Bghiggins/Ucdavis/Ech256/Jan - 2012
12 pages
SVD and PCA
No ratings yet
SVD and PCA
36 pages
MLF Combined
No ratings yet
MLF Combined
84 pages
Exercise 4.7, 4.9 Null Space, Rank of Matrix
No ratings yet
Exercise 4.7, 4.9 Null Space, Rank of Matrix
9 pages
B.Tech Syllabi - Branch-CSE, 2022 Regulations
No ratings yet
B.Tech Syllabi - Branch-CSE, 2022 Regulations
76 pages
14 BSC Maths CBCS Revised Syllabus Rev April 16
No ratings yet
14 BSC Maths CBCS Revised Syllabus Rev April 16
16 pages
Orthogonality and Vector Spaces
No ratings yet
Orthogonality and Vector Spaces
18 pages
Linear Algebra Elizabeth S Meckes Mark W Meckes Instant Download
No ratings yet
Linear Algebra Elizabeth S Meckes Mark W Meckes Instant Download
76 pages
01hybrid Gaussian-Cubic Radial Basis Functions For Scattered Data Interpolation
No ratings yet
01hybrid Gaussian-Cubic Radial Basis Functions For Scattered Data Interpolation
16 pages
CS771 IITK EndSem Solutions
100% (1)
CS771 IITK EndSem Solutions
8 pages
Linear Algebra Ques 2
No ratings yet
Linear Algebra Ques 2
4 pages
Lect-14 &15 (lsp1) - 2023
No ratings yet
Lect-14 &15 (lsp1) - 2023
42 pages
Methods For Ordinary Differential Equations: 5.1 Initial-Value Problems
No ratings yet
Methods For Ordinary Differential Equations: 5.1 Initial-Value Problems
20 pages
Formal Definition History Eigenvalues and Eigenvectors of Matrices
No ratings yet
Formal Definition History Eigenvalues and Eigenvectors of Matrices
18 pages
Inner Products: Wei-Ta Chu
No ratings yet
Inner Products: Wei-Ta Chu
44 pages
Linear Algebra - Module 1
No ratings yet
Linear Algebra - Module 1
22 pages
Linear Algebra - Friedberg - 4th Ed - Some Notions
0% (1)
Linear Algebra - Friedberg - 4th Ed - Some Notions
16 pages
1-1 LIN AL PYQs CHAPTERWISE TOPICWISE 2024
No ratings yet
1-1 LIN AL PYQs CHAPTERWISE TOPICWISE 2024
44 pages
LINEAR ALGEBRA AND VECTOR CALCULUS (GTU 2017) 4th Edition - Ebook PDF Instant Download
100% (1)
LINEAR ALGEBRA AND VECTOR CALCULUS (GTU 2017) 4th Edition - Ebook PDF Instant Download
80 pages
MTH215 Linear Algebra
No ratings yet
MTH215 Linear Algebra
2 pages
Solving Systems of Algebraic Equations: Daniel Lazard
No ratings yet
Solving Systems of Algebraic Equations: Daniel Lazard
27 pages
B.tech CSE CourseBook (21 25) WithAIML
No ratings yet
B.tech CSE CourseBook (21 25) WithAIML
144 pages
Cuda 2d
No ratings yet
Cuda 2d
8 pages
M2 - FDS
No ratings yet
M2 - FDS
20 pages
Syllabus M.A. M.sc. B.A and B.SC Mathematics
No ratings yet
Syllabus M.A. M.sc. B.A and B.SC Mathematics
26 pages
Math 3110: Linear Algebra: Spring 2014
No ratings yet
Math 3110: Linear Algebra: Spring 2014
3 pages