Fwd: a question about GWR4.0

Julia Koschinsky

unread,

Feb 23, 2015, 9:07:01 AM2/23/15

to openspa...@googlegroups.com, Ella...@unisa.edu.au

on behalf of Ella

---------- Forwarded message ----------
From: Ella Liu <Ella...@unisa.edu.au>
Date: 23 February 2015 at 01:52
Subject: a question about GWR4.0
To: "geoda...@asu.edu" <geoda...@asu.edu>

Dear Sir/Madam,

I met a problem when I used GWR4.0 software. I chose Logistic (binary) model. I cannot understand why the results showed “NaN”. Please find attached the data I used. It would be greatly appreciated if you could help me.

Best wishes,

Ella

The output result is listed below:

*****************************************************************************

Program began at 23/02/2015 5:21:33 PM

*****************************************************************************

Session: 2

Session control file: D:\New folder (2)\b.ctl

*****************************************************************************

Data filename: D:\New folder (2)\try2.csv

Number of areas/points: 200

Model settings---------------------------------

Model type: Logistic

Geographic kernel: adaptive bi-square

Method for optimal bandwidth search: Golden section search

Criterion for optimal bandwidth: AICc

Number of varying coefficients: 10

Number of fixed coefficients: 0

Modelling options---------------------------------

Standardisation of independent variables: On

Testing geographical variability of local coefficients: OFF

Local to Global Variable selection: OFF

Global to Local Variable selection: OFF

Prediction at non-regression points: OFF

Variable settings---------------------------------

Areal key is not specified

Easting (x-coord): field14 : X

Northing (y-coord): field15: Y

Cartesian coordinates: Euclidean distance

Dependent variable: field2: rural

Offset variable is not specified

Intercept: varying (Local) intercept

Independent variable with varying (Local) coefficient: field3: elevation

Independent variable with varying (Local) coefficient: field4: slope

Independent variable with varying (Local) coefficient: field5: precipitaion

Independent variable with varying (Local) coefficient: field6: dis_railway

Independent variable with varying (Local) coefficient: field7: dis_cbd

Independent variable with varying (Local) coefficient: field8: dis_coast

Independent variable with varying (Local) coefficient: field9: dis_road

Independent variable with varying (Local) coefficient: field10: popdensity

Independent variable with varying (Local) coefficient: field11: income

*****************************************************************************

Global regression result

*****************************************************************************

< Diagnostic information >

Number of parameters: 10

Deviance: NaN

Classic AIC: NaN

AICc: NaN

BIC/MDL: NaN

Percent deviance explained NaN

Variable Estimate Standard Error z(Est/SE) Exp(Est)

-------------------- --------------- --------------- --------------- ---------------

Intercept NaN NaN NaN NaN

elevation NaN NaN NaN NaN

slope NaN NaN NaN NaN

precipitaion NaN NaN NaN NaN

dis_railway NaN NaN NaN NaN

dis_cbd NaN NaN NaN NaN

dis_coast NaN NaN NaN NaN

dis_road NaN NaN NaN NaN

popdensity NaN NaN NaN NaN

income NaN NaN NaN NaN

*****************************************************************************

GWR (Geographically weighted regression) bandwidth selection

*****************************************************************************

Bandwidth search <golden section search>

Limits: 70, 200

Golden section search begins...

Initial values

pL Bandwidth: 104.073 Criterion: NaN

p1 Bandwidth: 140.714 Criterion: NaN

p2 Bandwidth: 163.359 Criterion: NaN

pU Bandwidth: 200.000 Criterion: NaN

Error in the initial weight calculation loop

Index was outside the bounds of the array.Best bandwidth size 0.000

Minimum AICc NaN

Error: GWR computation was failed: optional tests and/or model selection are cancelled.

*****************************************************************************

Program terminated at 23/02/2015 5:21:34 PM

--

************************
Julia Koschinsky, Ph.D.
Research Director
Associate Research Professor
Arizona State University
School of Geographical Sciences and Urban Planning
GeoDa Center for Geospatial Analysis and Computation
julia.ko...@asu.edu

http://geodacenter.asu.edu
http://www.facebook.com/geodacenter
http://twitter.com/GeoDaCenter

try2.csv

Julia Koschinsky

unread,

Feb 27, 2015, 8:25:14 PM2/27/15

to openspa...@googlegroups.com

try2.csv

Julia Koschinsky

unread,

Mar 11, 2015, 12:50:33 PM3/11/15

to openspa...@googlegroups.com, Ella...@unisa.edu.au

Response from my colleague below.

Julia

I checked the data. One issue is that the two independent variables "popdensity" and "income" only have two values, and if they are treated as binary variables, the two variables will have exactly the same value for each record. This will be a problem for any kind of regression, which has nothing to do with GWR.

Also, anther problem is the use of Euclidean distance while the coordinates are lat&long, which I think is not the essential reason causing the failure of the model.

Syahrul Effendi

unread,

May 27, 2019, 4:28:35 AM5/27/19

to Openspace List

Dear Sir/Madam,

I met a problem when I used GWR4.0 software. I chose Logistic (binary) model. I cannot understand why the results showed “Error: GWR computation was failed: optional tests and/or model selection are cancelled.”. Please find attached the data I used. It would be greatly appreciated if you could help me.

Best wishes,

Syahrul

The output result is listed below:

Program began at 27/05/2019 07.35.42

*****************************************************************************

Session:

Session control file: F:\acc.ctl

*****************************************************************************

Data filename: F:\data IPM se Sumatra.txt

Number of areas/points: 154

Model settings---------------------------------

Model type: Logistic

Geographic kernel: adaptive bi-square

Method for optimal bandwidth search: Golden section search

Criterion for optimal bandwidth: AICc

Number of varying coefficients: 1

Number of fixed coefficients: 4

Modelling options---------------------------------

Standardisation of independent variables: On

Testing geographical variability of local coefficients: On

Local to Global Variable selection: On

Global to Local Variable selection: On

Prediction at non-regression points: OFF

Variable settings---------------------------------

Areal key is not specified

Easting (x-coord): field3 : Long

Northing (y-coord): field2: Lat

Cartesian coordinates: Euclidean distance

Dependent variable: field1: Y

Offset variable is not specified

Intercept: varying (Local) intercept

Independent variable with fixed (Global) coefficient: field4: X1

Independent variable with fixed (Global) coefficient: field5: X2

Independent variable with fixed (Global) coefficient: field6: X3

Independent variable with fixed (Global) coefficient: field7: X4

*****************************************************************************

Global regression result

*****************************************************************************

< Diagnostic information >

Number of parameters: 5

Deviance: 116,992441

Classic AIC: 126,992441

AICc: 127,397846

BIC/MDL: 142,177204

Percent deviance explained 0,426550

Variable Estimate Standard Error z(Est/SE) Exp(Est)

-------------------- --------------- --------------- --------------- ---------------

Intercept -0,783788 0,243708 -3,216088 0,456673

X1 -0,088750 0,310790 -0,285562 0,915074

X2 0,420225 0,237095 1,772393 1,522305

X3 2,664963 0,459553 5,799028 14,367424

X4 0,169793 0,243606 0,697000 1,185060

*****************************************************************************

GWR (Geographically weighted regression) bandwidth selection

*****************************************************************************

Bandwidth search <golden section search>

Limits: 60, 154

Golden section search begins...

Initial values

pL Bandwidth: 84,637 Criterion: NaN

p1 Bandwidth: 111,132 Criterion: NaN

p2 Bandwidth: 127,506 Criterion: NaN

pU Bandwidth: 154,000 Criterion: NaN

Best bandwidth size 0,000

Minimum AICc NaN

Error: GWR computation was failed: optional tests and/or model selection are cancelled.

*****************************************************************************

Program terminated at 27/05/2019 07.35.54

Julia Koschinsky, Ph.D.
Research Director
Associate Research Professor
Arizona State University
School of Geographical Sciences and Urban Planning
GeoDa Center for Geospatial Analysis and Computation

julia.k...@asu.edu

data IPM se Sumatra.txt

Reply all

Reply to author

Forward