1997 Computstat Review

See
discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/2584514
Book and Software Review of "Yvan Pannatier:

Variowin --- Software for Spatial Data Analysis
in 2D"
Article · May 1997

Source: CiteSeer
CITATIONS READS
0 153
2 authors, including:
Jürgen Symanzik
Utah State University
118 PUBLICATIONS 920 CITATIONS
SEE PROFILE
All content following this page was uploaded by Jürgen Symanzik on 04 March 2013.
The user has requested enhancement of the downloaded file.

Book and Software Review of \Yvan Pannatier:
Variowin | Software for Spatial Data Analysis
in 2D"
Jurgen Symanzik and Noel Cressie
Department of Statistics, Iowa State University, Ames, IA 50011,
USA
[email protected], [email protected]
The book discussed in this review is one of the most recent additions
to Springer's series on\Statistics and Computing". In contrast to many
other books in that series, where multi{purpose statistical packages such as
S , PLUS R and XploRe are the matter of interest, Pannatier's book covers
a very specialized collection of four WindowsTM programs that can be used
for spatial data analysis and variogram modeling of irregularly located data
in two dimensions.
The 91 pages of the book are split into six chapters, appendix, bibliog-
raphy, and index. It includes a 3.5" diskette with programs containing the
latest version of Variowin, i. e., version 2.2.
The introduction in Chapter 1 gives an overview of the main features of
the program and summarizes the history of the software, originally developed
as part of Pannatier's Ph.D. research.
Chapter 2 is probably the most important part of the book for readers
with little time. It gives a quick start into the four programs included in
Variowin 2.2: Prevar2D (construction of a pair comparison le), Vario2D
with PCF (exploratory variography), Model (interactive variogram modeling),
and Grid Display (displaying grid les as pixel maps). Each page of this
S{PLUS is a registered trademark of MathSoft, Inc.
Windows is a registered trademark of Microsoft, Inc.
2
chapter contains screendumps from one of the four previously listed programs.
The text indicates which steps are required to obtain the window contents
visible in the screendumps. The required demonstration data used here is
provided on the accompanying diskette.
Chapter 3 is a very short introduction of pair comparison les (PCFs) and
describes how these are created using Prevar2D and what kind of problems
had to be overcome.
In Chapter 4, the interactive exploratory variography features that are
available in Vario2D with PCF are described. This program allows the calcu-
lation and display of (cross) h{scatterplots, (cross) variogram surfaces, direc-
tional (cross) variograms, and (cross) variogram clouds. Whenever one of the
options Calculate | Variogram Surface or Calculate | Directional
Variogram is selected, the following measures of spatial continuity can be
displayed: (cross) variograms, standardized (cross) variograms, (cross) co-
variances, (cross) correlograms, and (cross) madograms. Denitions of these
are also given in this chapter.
Pannatier uses (what we call)
^ij (h) =
X
1 N (h)(z (x ) , z (x + h)) (z (x ) , z (x + h));
2N (h) k=1 i k i k j k j k
where h is a separation vector, to calculate the experimental cross variogram

(p. 38). This is an estimator for
1
ij (h) = Cov((zi (s + h) , zi (s)); (zj (s + h) , zj (s))):
2
Even though 2ij (h) is traditionally recommended for the cross variogram
(Journel and Huijbregts, 1978, p. 324), it can be used only under the as-
sumption that cross covariances are symmetric in (h; ,h) (Journel and Hui-
jbregts, 1978, p. 326). However, there exists another denition of the cross
variogram, namely,
2 ij (h) = Var(zi (s + h) , zj (s));
which was rst considered by Clark, Basinger, and Harper (1989). It has been
shown in Ver Hoef and Cressie (1993, 1994) that 2 ij (h) does not depend on
the \symmetric{cross{covariance condition" and, moreover, linear prediction
based on 2 ij (h) is more ecient than linear prediction based on 2ij (h).
We recommend that fzi (xk )g and fzj (xk )g be rst standardized so that z i =
z j = 0 and s2i = s2j = 1 and then, that 2 ij (h) be estimated by
XN (h)
2^ ij (h) = N 1(h) (zi (xk + h) , zj (xk ))2 :
k=1
While all other programs of the Variowin 2.2 package are almost self{
explanatory, the program Model (used for interactive variogram modeling)
3
and its input window require an explicit description, given in Chapter 5.

Five variogram models are supported by the software: pure nugget{eect
models, spherical models, exponential models, Gaussian models, and power
models.
Chapter 6 is intended for the advanced reader/user of the software since
it discusses the format of data les used and generated by the four programs
in the Variowin 2.2 package.
An appendix on geostatistical concepts provides some of the necessary
theoretical background for the work possible with the Variowin 2.2 package.
However, it might be very helpful for the reader to look additionally at some
of the standard references on geostatistics, e. g., Isaaks and Srivastava (1989)
and Cressie (1993). A bibliography and an index conclude the book.
The Variowin 2.2 software runs on any PC with a 386 processor and above
(running WindowsTM 3.x, 95, or NT), a coprocessor, and approximately 2MB
hard disk capacity. The four programs are mostly operated through the
mouse while function keys serve as accelerators. Unfortunately, the split of
the software into four (dependent) programs is a very user{unfriendly fea-
ture. In a typical application, the program Prevar2D has to be started rst
to create the PCF. Then, Vario2D with PCF is started. It would have been
more convenient if a single option, e. g., in the \File" menu of Vario2D with
PCF, provided the functionality of Prevar2D. Similarly, after the interactive
variogram modeling has been done in Model, the user explicitly has to create
external les and start the program Grid Display to display pixel maps. We
would recommend that a future version of Variowin joins the four programs
into one, thus removing the burden from the user to (re{)start dierent pro-
grams and to create and handle external les.
Two more things should be mentioned about the user interface: The
way in which multiple subwindows within the Variowin main window can be
customized and handled is very convenient and helpful. Otherwise, on{line
help for the features and options of the programs is almost non{existent.
As far as the features on exploratory variography (provided by Vario2D
with PCF) and variogram modeling (provided by Model) are concerned, Var-
iowin is denitively one of the most sophisticated programs that currently
provide this type of application. Therefore, the book should be considered as
a users' guide to Variowin that directly addresses exploratory geostatistical
data analysis. Without the software, the book has very little intrinsic merit.
One further critical comment about the scope of the software should be
made. Recently, many links between software packages have been developed
where the statistical application operates directly on data extracted from
a geographic information system (GIS), e. g., S{PLUS for ARC/INFO by
MathSoft, the link between SpaceStat and ArcViewTM (Anselin & Bao 1996),
and the link between XGobi and ArcViewTM (Cook, Majure, Symanzik &
Cressie 1996); however, the Variowin 2.2 package can be used only as stand{
TM ArcView is a trademark of Environmental Systems Research Institute, Inc.
4
alone software. The new links between software packages referred to above
allow one to modify displays in the statistical package while the sample map
in the GIS, that also displays many other types of geographic information, is
concomitantly modied (and vice versa). Variowin only allows simultaneous
modications of sample maps and plots in its window but, unfortunately, no
additional geographic information can be displayed.
In the long run, Variowin also has to compete against S+SpatialStats from
MathSoft. Many researchers are already familar with the basic S , PLUS R
software and, therefore, may not be willing to learn how to use a new pack-
age. In general, S+SpatialStats oers a much broader functionality for the
exploration and modeling of spatially correlated data, including some (but
not all) of the features provided by Variowin. However, it should be left to
the reader to decide if he or she is interested in a comprehensive software
package for the analysis of spatial data (provided by S+SpatialStats) or in
the very specialized and sophisticated exploratory variography and variogram
modeling for geostatistical data (provided by Variowin).
References
Anselin, L. & Bao, S. (1996), Exploratory Spatial Data Analysis Linking
SpaceStat and ArcView, Technical Report 9618, West Virginia Univer-
sity, Morgantown, WV.
Clark, I., Basinger, K. L. & Harper, W. V. (1989), MUCK | A Novel
Approach to Co{Kriging, in B. E. Buxton, ed., `Proceedings of the
Conference on Geostatistical, Sensitivity, and Uncertainty Methods for
Ground{Water Flow and Radionuclide Transport Modeling', Battelle
Press, Columbus, OH, pp. 473{493.
Cook, D., Majure, J. J., Symanzik, J. & Cressie, N. (1996), `Dynamic Graph-
ics in a GIS: Exploring and Analyzing Multivariate Spatial Data Using
Linked Software', Computational Statistics 11(4), 467{480.
Cressie, N. A. C. (1993), Statistics for Spatial Data (revised edition), Wiley,
New York, NY.
Isaaks, E. H. & Srivastava, R. M. (1989), Applied Geostatistics, Oxford Uni-
versity Press, New York, NY.
Journel, A. G. & Huijbregts, C. J. (1978), Mining Geostatistics, Academic
Press, London.
Ver Hoef, J. M. & Cressie, N. (1993), `Multivariable Spatial Prediction',
Mathematical Geology 25(2), 219{240.
Ver Hoef, J. M. & Cressie, N. (1994), `Errata: Multivariable Spatial Predic-
tion', Mathematical Geology 26(2), 273{275.
View publication stats

1997 Computstat Review

Uploaded by

Copyright:

Available Formats

1997 Computstat Review

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

1997 Computstat Review

Uploaded by

Copyright:

Available Formats

See

Book and Software Review of "Yvan Pannatier:

Article · May 1997

The user has requested enhancement of the downloaded file.

where h is a separation vector, to calculate the experimental cross variogram

and its input window require an explicit description, given in Chapter 5.

View publication stats

You might also like