----------------------------Original message----------------------------
Thanks to everyone who sent me replies to my query (over 80!). Here is
a summary of what I learned.
AN INTERNET SURVEY OF COURSES
AND TEXTS ON SPATIAL DATA
ANALYSIS
Marshall M.A. Feldman
Associate Professor
CPAD Working Paper 95-02
Graduate Curriculum in Community Planning and Area Development
204 Rodman Hall
The University of Rhode Island
Kingston, RI 02881-0815
Phone: 401/792-5953
FAX: 401/792-4395
Internet: [log in to unmask]
AN INTERNET SURVEY OF COURSES AND TEXTS ON
SPATIAL DATA ANALYSIS
On May 3, 1995 I queried several Internet lists and newsgroups about courses and
textbooks for spatial data analysis. The query generated considerable interest,
with over
eighty replies within a week. This is a summary of those replies and is
organized as
follows:
1 The Original Query
1.1 Lists to Which Queries Were Sent
1.2 Text of the Original Query
2 Descriptions of Courses on the Subject
2.1 Courses Emphasizing Statistics
2.2 GIS-Based Courses
2.3 Courses Combining GIS and Statistics
2.4 Planning Methods Courses
2.5 Courses with Other Emphasis or Emphasis Unknown
3 Computer Software
3.1 Academic Atlas
3.2 ARC/INFO
3.3 ArcView2
3.4 Atlas GIS and Atlas Pro
3.5 Census Data
3.6 EpiInfo/EpiMap
3.7 Geo-EAS 1.2.1
3.8 GEOEAZ (Is this Geo-EAS?)
3.9 GEO-PACK
3.10 GeoQuery
3.11 GeoSim
3.12 GRASS
3.13 GS+
3.14 GSLIB
3.15 IDRISI
3.16 Landview
3.17 MapInfo
3.18 MapViewer
3.19 Minitab for Windows
3.20 Pointstat
3.21 PMAP
3.22 PVWave
3.23 R programs
3.24 REGARD
3.25 Regional Analysis System (RAS)
3.26 S+
3.27 SAS
3.28 SPACESTAT
3.29 SPANS
3.30 SPSS for Windows and Add-Ons
3.31 STAC
3.32 Statview
3.33 SURFER for Windows (Golden Software).
3.34 SYSTAT
3.35 USTAT - Jerry Schneider ([log in to unmask])
3.36 Other
4 Miscellaneous Comments
4.1 On the Utility of Such A Course and How to Package It
4.2 On How The Course Fits into the Planning Curriculum
4.3 Other
5 Results of a WebCrawler Search
6 Texts on the Subject and References
6.1 General Comments
6.2 References
1 The Original Query
1.1 Lists to Which Queries Were Sent
The query was posted to these lists:
Bmdp-l: BMDP List
Envst-l: Environmental Studies List
ESRI-l: Environmental Systems Research Institute List
Geograph: Geography List
GIS-l: Geographic Information Systems List
H-urban: Urban History List
Maps-l: Map List
Methods: Teaching Social Science Methods List
Planet: Planning Educators List
Regsc-l: Regional Science List
S-news: S and S+ News
SAS-l: SAS List
SASPAC-l: SAS Public Access Consortium
SPSSX-l: SPSSX List
Stat-l: Statistics Consulting List
Urban-l: Urban Planning Discussion List
Urbgeog: Urban Geography List
1.2 Text of the Original Query
I'm sorry about the cross-posting, but this is a fishing trip and I want to
cast the net as widely as possible. If there is sufficient response, I'll
summarize and re-post to the respective lists. Maybe this is even a FAQ
on some lists.
I am thinking about designing a course on spatial data analysis for
planning students with an emphasis on social, economic, demographic, and
perhaps some environmental data. Its relationship to spatial statistics
would be roughly the same as a course on data analysis' relationship to
aspatial statistics. In other words, there would be more emphasis on
organizing and accessing data, simple exploratory and descriptive data
analysis and presentation, plus some dose of more conventional topics in
spatial statistics (e.g. spatial autocorrelation). The course would be offered
primarily to second-semester graduate students in community planning who
have had one introductory statistics course and a course on research
methods.
Since this might substitute for one of our current planning methods
courses, it might also have to spend some time on things like population
projections, census data, etc.
Does anyone out there teach a course like this? Do any of you have
ideas or suggestions about such a course? Two things of particular
concern are textbooks and software. Can anyone suggest a textbook that
would be appropriate? What about software with general purpose
statistical capabilities and the ability to make decent, presentation-quality
analytical maps? One of the things I've learned the hard way is that any
software like this would have to be extremely user-friendly or else the
software becomes the focus of the course. While bright, some of these
students have minimal computer experience, and I can't imagine using an
algorithmic package in this course: this rules out S+, SAS (even with
ASSIST, SAS GRAPH is way too complicated), and ARC/INFO. Ideally,
the software should be cheap and run on fairly basic PC's (but this may be
wishful thinking).
Please let me know what you think, and if there's sufficient response
I'll summarize and post to this list. Thanks for your time.
Well, there certainly was sufficient response. The subject appears to interest
many people
from diverse disciplines. Replies ranged from, "let me know what you find out"
to "I've
been doing this for several years." Some replies discussed fairly elementary
approaches,
others strongly advised that such a course could only be taught to students with
several
semesters of aspatial statistics under their belts.
2 Descriptions of Courses on the Subject
2.1 Courses Emphasizing Statistics
2.1.1 Lynn Rosentrater a graduate student in the Department of
Geography,
University of Oregon ([log in to unmask]) took a course called Geographic
Data
Analysis using Chapman and Monroe (1993) and Minitab for Windows. She also TA'd
a course called Statistics for Social Scientists which uses SPSS and Norusis
(1990).
2.1.2 Dartmouth College (Adrian Bailey)
Megan Blake also recommends Dr. Adrian Bailey ([log in to unmask]) at
Dartmouth who used spatial autocorrelation considerably and she believes has
also has
taught a statistical methods class like this.
2.1.3 Duke University
Richard Smith taught a course in spatial statistics at Duke a couple years ago.
It was a
little more theoretical than what you have described, though most of the
students were in
an environmental science curriculum. You might want to contact him. I don't
know his
Cambridge e-mail address, but he still has one at UNC, which I believe he checks
now
and then: [log in to unmask] -- Patrick Crockett, [log in to unmask]
2.1.4 Edith Cowan University (Lyn Bloom, [log in to unmask])
I do run such a course (called MAT5104 Spatial Data Analysis) here at Edith
Cowan
University as part of our MSc (Mathematics and Planning) degree. At present I
am using
Upton & Fingleton (1985, vol. 1) covering most of the material in Chapters 1-4
with
supplementary material on spatial EDA and geostatistics from Cressie (1991) and
Isaaks
and Srivastava (n.d.).
Software is a problem but I have been using MINITAB 10 for Windows (with
appropriate macros) and also the public domain Geostatistics package GEOEAS.
The
students also have access to IDRISI and MapInfo.
Here is the course content:
1 Exploratory data analysis of spatial data.
2 The Identification of Pattern: alternative patterns, quadrat counts,
distance methods,
spatial mapping methods.
3 The Estimation of Spatial Intensity: quadrat methods, line transects,
distance methods,
areal methods.
4 Spatial Autocorrelation: randomization, Monte Carlo approach, normal
approximation,
join-count statistics, Moran's I statistic, Geary's c statistics,
correlograms.
5 Inter-type Relations: bivariate point patterns, analysis using quadrats,
analysis using
transects, distance analysis, spatial rank correlation.
6. Spatial Prediction: triangulation, h-scatterplots, variograms, ordinary
kriging.
2.1.5 University of Idaho (Scott Morris)
The Geography Department at the University of Idaho offers a 300-level course
called
"spatial analysis." It is taught by Scott Morris and uses Clark and Hosking.
The course
is part of the geography core and has a 200-level statistics class as a
prerequisite. The
class goes through the glm with some emphasis on cluster and factor analysis.
2.1.6 Indiana University (Daniel Knudsen)
Daniel Knudsen ([log in to unmask]) at Indiana University (Geography)
teaches
a similar course. I gather he uses Clark and Hosking as well as Hanushek and
Jackson.
2.1.7 Iowa State (Noel Cressie)
Noel Cressie teaches a course like this at Iowa State. He was visiting Ohio
State and
teaching a seminar on Spatial Statistics. The textbook written by him would be
good for
such a course. -- John P Lawrence.
2.1.8 Michigan State University (Bruce Wm. Pigozzi
[log in to unmask])
I teach a course somewhat like what you describe (Quantitative Methods for
Geographers
and Planners). It attracts about 50-60 students each fall semester. Recent
texts: Earickson
and Harlin (1994), McGrew and Monroe (1993), and Clark and Hosking (1986). I've
been most satisfied with Earickson and Harlin for the basic course (there's also
a
multivariate oriented second course) in a semester format. Bruce uses SYSTAT in
this
course.
2.1.9 North Carolina State, Raleigh (G. Peter Y. Clarke
[log in to unmask]
.za)
In 1991 I spent a sabbatical at NC State, Raleigh in the Statistics Dept. I
offered a
"special" course on spatial statistics, comprising 1 lecture per week for 9
weeks
plus labs. The audience was mostly post-grad and staff with a fair sprinkling
of forestry
students, so I couldn't make it too mathematical. Essentially, I covered mostly
various
forms of kriging and co-kriging. I was unable to find a suitable text, and made
up my
own notes.
For labs I used GEO-EAS , which worked quite well. Subsequently, the course
has
been taught by Marcia Gumpertz ([log in to unmask]), and she has been using
GSLIB in preference to GEO-EAS because the latter package fits in with the UNIX
environment. Both GEO-EAS and GSLIB are free-ware.
2.1.10 UCLA (Cindy Fan)
Eric Stein ([log in to unmask]) writes:
While working on my doctoral degree in environmental science and engineering I
took
a course on spatial statistics ... in the Geography Department, but geared
toward
application of the methods to each student's personal data. The class covered
descriptive
as well as quantitative methods. Overall a pretty good class.
The professor was Cindy Fan; maybe you could try contacting her for information
on the
course. I don't have an email address for her, but she can be reached at the
UCLA Dept.
of Geography, 1255 Bunch Hall, Los Angeles Ca 90024.
2.1.11 University of Iowa (Dale Zimmerman, [log in to unmask])
Dale Zimmerman, in Statistics, uses S+ when teaching spatial statistics and
recommends
Haining (1990) as a text.
2.1.12 University of Montreal (Universite de Montreal, Pierre Legendre,
[log in to unmask])
I have been teaching a course on spatial statistics for ecologists for several
years. I have
also had the opportunity to give it (whole, or parts of) in various other
countries, in
French or in English (France, Belgium, Italy, Norway, Spain...). I will send
you by paper
mail the table of contents of the one I gave in Spain last fall. Hope you find
some useful
ideas in it.
Clientele: Typically, graduate students in ecology, with some outliers coming
from
geography, engineering, etc.
Textbooks: None at the moment. I agree with you that we should not use textbooks
of
spatial statistics written by statisticians for this type of teaching. So, I use
a series of
published papers -- many from my own lab -- as well as chapters of the book that
I am
in the process of writing (Numerical Ecology, 2nd English edition). I have a
section on
spatial autocorrelation; one on permutation testing; and I will have a whole fat
chapter on
spatial analysis (an embryo of it has already published in my 1984 book).
Pierre also uses the R package and SURFER (see below) for software.
2.1.13 University of Rhode Island (Alan Humphrey, Management Science and
Information Systems, [log in to unmask])
There is a new program for the PC/Windows that is suppose to be very easy to
use. It
constructs chloroplethic, dot, and column maps and also has an addmatcher
included. The
cost is about $400. I'm getting a demo and additional information that I would
be happy
to share with you.
I have taught courses in the distant past that touch on spatial analysis and
I'm sure we
would have some marketing students here that would be most interested in such
and
offering. Keep in touch.
2.1.14 University of West Virginia (Luc Anselin)
I have taught such a course several times (though not to community planning
persons) and
have a few suggestions:
- if one wants to avoid a simple cookbook approach, more than a single
stats course
is needed as prerequisite; most of the spatial stats is actually more
difficult than
non-spatial stats
- there is no software that does both mapping and graphics (in a "user
friendly"
way) and spatial stats. In fact, as far as I know there are no
introductory spatial
stats packages that do not involve substantial programming (a la S-Plus,
for
example). The closest to your need would be my Spacestat package (blurb
below), but it probably requires more background. It is used in a
teaching
environment in a number of places, including Carnegie Mellon (in a
Masters level
class), San Diego State (geography upper level), Kentucky (upper level
geography)
and some other places.
- my suggestion would be to use an "easy" mapping package like Arc/View in
combination with a spatial data analysis package (such as SpaceStat),
which would
require some computer literacy, but not an inordinate amount. I am
teaching a 4
week intensive summer course like that at ICPSR (University of Michigan)
this
summer. (Note: Information on the ICPSR course, including syllabus and
reading
list can be obtained at
gopher://gopher.icpsr.umich.edu:70/11/Summer_Program).
2.2 GIS-Based Courses.
2.2.1 Michael Goodchild of NCGIA, UC at Santa Barbara teaches a
"Spatial Analysis Using GIS" Seminar.
2.2.2 University of Wisconsin-Milwaukee (Bill Huxhold [log in to unmask])
Teaches an introductory GIS course. He sent me his lab manual which involves
exercises
on data and spatial analysis representing typical local government functions
(parcel data
maintenance, geocoding address-based parcel data, building permit updating,
census data,
data aggregation on housing units and land use, residential density analysis,
and flood
plain overlay and reporting). The students first work with the data in a manual
mode (he
sent the Instructor's Version) and then move on to use PC Arc/Info to do the
same work
and more of the GIS functions. The manual is 250 pages in length and will be
published
by Oxford University Press next year.
I looked at the manual, and it involves tasks typical of using GIS for
database queries.
For example, students are asked to identify all parcels in a flood plain by
doing a spatial
union.
2.3 Courses Combining GIS and Statistics
2.3.1 Art Getis ([log in to unmask])
Art Getis teaches a course to students who are mainly interested in GIS and
environmental
problems and updates it from year to year to incorporate the latest software
(and
thinking). His students are usually third semester graduates who have had a
good course
on statistics (multivariate analysis).
He uses no textbook; the course readings are taken from about twenty books,
several
articles, and several software manuals. The main books are those by Haining
(1990),
Upton and Fingleton, Anselin, Boots and Getis, Cliff and Ord, Davis, and
Cressie.
Software: S+ (it can be used with novices if one outlines the basic idea of it),
GIS
software, GEOEAS (I will be using new software on Kriging this year), SpaceStat,
and
my own Fortran programs on spatial pattern analysis.
2.3.2 Clemson University (Shuming Bao, [log in to unmask])
In the Department of AG/Applied Economics, Professor Bao teaches "Introduction
to GIS
with ARC/INFO and Spatial Statistics" graduate students. Newly developed
spatial
statistics such as Getis and Ord's G statistics and Luc Anselin' LISA (Local
Indicator of
Spatial Association) can be found in papers. I doubt if there is good software.
All those
spatial statistics are associated with some kinds of spatial weight that can be
derived from
GIS. The new ARC/INFO (7.0) might have G statistics in GRID model. The
software,
SpaceStat, developed by Luc Anselin can calculate G and LISA, which is totally
independent from GIS packages GIS packages.
2.3.3 Colorado State University
Dr. Robin Reich teaches a course on Quantitative Spatial Analysis here at
Colorado State
University, Forest Department, College of Natural Resources. -- Rodney Howe
([log in to unmask])
2.3.4 MIT (Joe Ferreira [log in to unmask])
Joe's TA (Alicia L Allen [log in to unmask]) Joe generally spends the first half of
the term
teaching the software and the second half doing group projects. This term we
used
Arc/Info and ArcView2. ArcView2 is enough of a user-friendly interface that it
can be
learned fairly quickly. The biggest problem is that you need to use something
like Oracle
to do querying and analysis, unfortunately this can become very expensive.
The course is given 2nd term to grad students and it is also an
undergraduate institute
lab, so there are some undergrads in it. However, at MIT the undergrads tend to
be fairly
computer-proficient. Also, the graduate students take a short course that
introduces them
to GIS and relational databases during our January interim session, so they have
some
familiarity as well. This is a required course for all Masters of City Planning
students at
MIT.
In previous years they used MapInfo, which is fairly simple for anyone to
learn the
basics of. Alicia once TA'ed a class of foreign planners and they picked it up
without
too much trouble.
The text is mainly a reader of articles that Joe collected, as well as
Huxley (no ref --
do you mean Huxhold?).
2.3.5 Ned Levine, UCLA and University of Hawaii
([log in to unmask]
du)
At UCLA, I taught a GIS course that had a strong analytical orientation; this
semester,
I'm teaching one at the University of Hawaii that is even more analytic. In
addition, I'm
writing a review of spatial statistics software for JAPA and have been looking
at a lot of
programs, texts, etc. My impression about integrating spatial statistics into
GIS classes
is that it would be fine for a second course, but not a first one. It takes
almost an entire
semester to make students feel comfortable with the current crop of spatial
analytical
routines in most GIS programs (ATLAS, ArcInfo); these routines are not really
statistical,
but are database-like operations (splitting, unioning, etc.). It would take
another semester
or so for students to learn enough about spatial statistics and to learn how to
link the
existing programs up with the GIS systems. Most of the spatial stats
programs/routines
are 'home-made' affairs and take a while to learn the idiosyncracies. Further,
most do not
hook up directly with a GIS but have to be run as a separate program. The one
exception
is the S+GisLink from StatSci, which hooks up S-Plus to ArcInfo; however, it
only works
for ArcINfo versions 6.1 or lower. We have 7.0 here and while I can run S-Plus
commands from the Arc command line, I can't transfer data between the two.
2.4 Planning Methods Courses
2.4.1 Cal Poly, San Luis Obispo
Mike Multah teaches a similar course called CRP 315, Planning and Economic
Analyses
(or similar title). Call Mike at 805-541-3848. -- Barbara Sommer,
[log in to unmask]
edu
2.4.2 The University of Cincinnati (Johanna W. Looye,
[log in to unmask])
I teach planning methods at the University of Cincinnati. One of my
graduate-level
courses ("Planning Methods") uses R. Klosterman's book Community Analysis and
Planning Techniques, but it isn't very spatial. I teach population projection,
cohort-
projection, economic base analysis, index numbers, and other "tried and true"
techniques.
Klosterman and others have a book out that uses spreadsheets, Spreadsheet Models
for
Urban and Regional Analysis, but I have never used it in a classroom. I find it
too
"black-boxy" for my students, who would simply type in numbers and read output
without
necessarily understanding the calculations involved. I use the first book and
have them
construct their own spreadsheets.
Another course I teach is called "Methods in Economic Development Planning"
but
I myself call it my "census course." I organize the quarter (10 weeks) around
producing
a regional compendium organized on Losch's population-location-activity model.
The
students work almost exclusively with spreadsheets to produce joint and
conditional
probability tables for a large variety of variables. They also learn to use
index numbers,
Lorenz curves and Gini coefficients, and other techniques. I, myself, provide
them tables
for the Cincinnati PMSA, by county, and then I divide the class into two teams
who
produce tables for two other MSAs in the US. The underlying theme is growth and
decline (so this year we're doing Raleigh-Durham-Chapel Hill and
Albany-Schenectady-
Troy) and the presence is wealth and poverty in each. The take-home final asks
students
to write about how the two cities shape up compared to a few articles we read
about the
national/regional trends.
I don't actually put anything on maps but it is geography based (county
level), so (and
I plan to talk to my colleagues who teach GIS) I think it's sort of pre-map
analysis of
data. A good book, though it works at the census tract level rather than the
county level,
is Dowell Myers' Analysis with Local Census Data. I also use Bendavid-Val's
Regional
and Local Economic Analysis for Practitioners which covers things like location
quotients,
distribution quotients, shift-share analysis.
I have a feeling we could have a long conversation about this course, so
I'll stop here
and let you respond (if you wish).
2.4.2.1 My Comments
I also teach straight planning methods courses like Johanna's, and I have used
both
Klosterman's Community Analysis and Planning Techniques and Meyer's Analysis
with
Local Census Data books. When using Klosterman, I also had the students make
their
own spreadsheets rather than give them canned ones.
While I agree that canned spreadsheets are too black-boxy, I do not feel
having
students do their own spreadsheets adequately addresses the problem. For one
thing, in
our curriculum students who do not know spreadsheets have to learn them while
they are
taking this course. We always have some students who have not seen math since
high
school or a computer ever, and learning both quantitative methods and
spreadsheets
simultaneously proves "challenging" to put it mildly. Second, while
Klosterman's book
could be improved by being more flashy -- Myers is better here -- the techniques
themselves leave much to be desired. Even if one knows spreadsheets, I'm not
sure that
mechanical application of codified methods, e.g. cohort population models, is
any less of
a black box for most students.
Perhaps this is unique to our curriculum. Klosterman covers three bodies of
methods:
trend extrapolation, population projection, and economic base analysis. All
three are used
for predicting future patterns. Yet our students have no prior exposure to
demographic
theory or urban economics, so I don't think learning these techniques as
"methods"
overcomes the black-box syndrome. The same is true for the trend extrapolation
techniques, which are purely mathematical. Without calculus and a whole lot of
social
theory and philosophy of science, it's hard to see how students can "properly"
apply trend
extrapolation techniques to social phenomena. For example, what does chaos
theory do
to our belief in the accuracy or appropriateness of projections?
For this reason, I went over to Myers this past year. My goal is not to
give students
a bag of "tools," although many of them say that's exactly what they want,
because I
think such bags actually get in the way of understanding. Instead, I want to
increase their
"dexterity" with quantitative data so that, for example, students can face an
unfamiliar
situation in which quantitative analysis might be called for and do something
intelligent.
Myers seems better suited for this purpose since it largely covers several ad
hoc analyses
of issues that come up when dealing with census data. Yet students seem
unsettled by
this, believing they will not know what to do when they get a job. To draw an
analogy
with art, they seem to want to know the rules for drawing portraits or
landscapes rather
than how to be artists.
Although I do not teach a course on methods in economic development
planning, I
have thought a good deal about such a course, and I do teach economic
development
planning. I have sometimes used Bendavid-Val's book, but if I were to teach
this course
I would put more emphasis on the analysis of local taxes.
Here again, I think there are serious problems teaching "methods" divorced
from
theory. This, of course, betrays my methodological and theoretical
dispositions.
Methodologically, I am inclined towards post-positivist theoretical realism,
which rejects
the notion that there is a single "scientific method" divorced from specific
contexts. In
this view, all data and methods are "theory laden," so teaching methods divorced
from
theory is basically a subtle form of dogma. Theoretically, I largely reject
neoclassical
approaches to economics and liberal approaches to economic development policy or
understanding the state. Instead, I am most comfortable with the "regulation
approach"
as a way of understanding these phenomena. So, while I teach and even sometimes
use
things like location quotients or economic base analysis, I am quick to point
out to
students that I think they're often pretty silly tools. As an example, I point
out how off-
base these techniques would have been had we applied them in the 1940's to what
we
now call Silicon Valley: they do not do a good job of either predicting or
explaining what
happened. Needless to say, students find this disquieting since they wonder why
they had
to go to the trouble learning the techniques and what to make of their "bag of
tools."
Johanna is right. This is the basis for a MUCH larger discussion. Maybe at
ACSP?
2.4.3 Tufts University
According to Melissa Twining Davis ([log in to unmask]) the Department of
Urban and Environmental Policy offers Quantitative Reasoning for Policy and
Planning.
The course description reads:
Range of methodologies and approaches involved in the analysis and
interpretation
of quantitative data. Emphasis on practical and decision-making aspects of
quantitative methodologies rather than research aspects. Descriptive
univariate
statistics, the interpretation of multivariate models, and a review of
commonly
used methods in decision analysis. Application to problems of program
evaluation
and planning. Introduction to and use of microcomputer and mainframe
computer
in decision making.
2.4.3.1 My Comment
What differentiates the "practical and decision-making aspects" of any
methodology from
the "ideological and non-critical" aspects? In 1617 Martin Luther posted a list
of
philosophical theses on the door to a church, and this is widely regarded as
marking the
beginning of the Reformation. Since it changed the world far more than most
land use
plans, I would say it was very "practical," but I don't know how one could just
look at
the substantive content of Luthor's posting and make this judgement.
2.4.4 Virginia Tech
I've been teaching one for about ten years and would be happy to share any
materials if
you like. The core of the course has been statistics, but I also spent time on
forecast
methods, some spatial analysis, and presentation graphics. We are changing
aspects of
our methods sequence next year so it will be somewhat different next year. --
Jim
Bohland, ([log in to unmask])
2.5 Courses with Other Emphasis or Emphasis Unknown
2.5.1 Hunter College
Paul Wood ([log in to unmask]) writes:
A number of professors teach a whole host of courses related to Spatial Data
Analysis at
Hunter College--CUNY in the Dept. of Geology and Geography. The Department has
a
Spatial Analysis Laboratory with some very nice hardware (Sparc Sun Stations)
and lots
of software. Jeff Osleeb--used to be Chair of the Dept.-- has tried for years
to put
together a PhD program in Environmental Science with an emphasis in GIS and
modeling.
Names include:
Dr. Jeff Osleeb--Spatial Analysis, Transportation Modeling
Dr. Sean Ahern--GIS
Tom Walter--Director of the Spatial Analysis Lab
Hunter College
Dept. of Geo. & Geog.
North Bldg, Rm 1006.
695 Park Ave.
NY, NY 10021
(212) 772-5265
Tony Grande is the Dept. Secry.
2.5.2 Ohio State
Gail Gill <[log in to unmask]> writes:
Yes, it (the course) has been taught at Ohio State University, by Dr. Joe
Verducci. I don't
remember his e -mail, but Dr. Saul Blumenthal (who also teaches there) might be
able to
give it to you. His address is [log in to unmask]
2.5.3 University of Maine
Kate Beard at the University of Maine is teaching a course with the same-ish
title this
next fall. You might want to talk to her. Her e-mail is
[log in to unmask]
-- Jeff Hepinstall ([log in to unmask])
3 Computer Software
3.1 Academic Atlas
Recently I acquired a program called Analytic Atlas which does thematic maps
with
colors or 3-D wire frames, cross-plots of variables, etc.. It is extremely easy
to use,
though not terribly powerful. It comes with a companion program, AAMaker which
allows users to construct maps from scratch, from PCX files and download
boundaries and
data from Tiger files. This component is a little more complex, but manageable.
The
neat thing is that the whole package is public domain and fits on one 720k
diskette. I may
be presenting it at ACSP this fall, but sending me a disk gets anyone a copy
now. I like
the idea of being able to turn over data by census tract or block or county to
decision
makers along with the analytic software so that they can generate their own
analyses,
though I haven't tried it in practice. (JIM FISHER: [log in to unmask])
3.2 ARC/INFO
3.2.1 Gary M. Greenberg ([log in to unmask])
During my Master of Arts in Urban & Regional Planning education, I took quant.
methods, and 3 GIS courses. UF's URP has a fine research facility called the
GeoPlan
Center, wherein Arc/Info is taught using RS6000 Workstations. Drs. John
Alexander and
Paul Zwick are the Director and Co-director of GeoPlan. They can be contacted
at
904-392-0997.
The projects we worked on are real world, such as modelling facility siting
using
relevant data <population statistics, pop. projections, permit approvals, info
garnered from
planning documents, property appraisers' data, and the like>.
Don't rule out Arc/Info! It is a dominant GIS force, and it's not that tough
to grasp.
I was a computer neophyte when I started my Masters!!! Now, i m a prgmr ;-)
Small
GIS tools don't really suffice in a planner's toolkit. The issues require a
tool which will
be able to handle whatever you throw it, and that means being able to create a
model on
top of the standard framework: which is why tools like Arc/Info and Intergraph
are
dominators. If the students won't learn to manage the more complex tools <SAS,
Splus,
whatever> then they won't be prime candidates in the search to implement what
they've
learned, provided they can communicate well.
3.2.2 [log in to unmask]
Avoid all ESRI products, no matter what they tell you in the brochures, these
are for the
technicians. Even the "user friendly" ones are hard work.
3.3 ArcView2
From [log in to unmask] at the University of North Carolina: I have
been
hearing wonderful things about ArcView2 from our City & Regional Planning Dept -
they
say this product does in 10-15 minutes what it takes a student to do in several
hours using
ArcInfo. Sounds so good we are going to get a copy this summer for our lab. I
know
they have educational pricing, but I don't know what it is.
3.4 Atlas GIS and Atlas Pro
3.4.1 Clinton J. Andrews ([log in to unmask])The projects we worked
on
are real world, such as
Teaches the spatial analysis component of the Woodrow Wilson School's methods
course
modelling facility siting using relevant data <population statistics, using
Atlas GIS
(Strategic Mapping, tel 408-970-9600), which is much more user friendly than
ArcInfo,pop. projections, permit approvals, info garnered from planning
documents, and
runs on our networked PCs (486's). The tutorial comes with some good data sets
and
property appraisers' data, and the like>. I've added other New Jersey-specific
ones for
their problem sets. The MPA students seem to require about 8 hours of hands-on
time
to become comfortable with the program, after which the fun stuff can begin.Drs.
John
Alexander and Paul Zwick are the Director and Co-director of GeoPlan. They can
be
contacted at 904-392-0997.
3.4.2 John Cook ([log in to unmask])
AtlasPro could be pretty easy to run on not-so-state-of-the-art desktop
machines, and it
can do some basic analytical stuff. Atlas GIS is now much more affordable, and
the
Windows version is pretty user friendly. It has even greater analytical
capability.
3.4.3 Barbara Sommer, [log in to unmask]
ATLAS GIS is easier than ARC/INFO, but still may be too much. Don't know about
presentation quality.
3.4.4 [log in to unmask]
Avoid commercial packages like Atlas GIS and MapInfo. While these are "sexy",
and
produce good output, they tend, like many commercial products, to deal with
theoretical
problems by ignoring them. If the system goes "bong" and some stuff lights up,
you've
probably just skipped over a major problem of auto-correlation or made a totally
unjustifiable extrapolation.
3.5 Census Data
John Cook ([log in to unmask]) says the Bureau of the Census is offering a
pretty nice
county-level package on CD-ROM, if that is a possibility. Otherwise there are
the
standard things you can get from them in hard copy, and some in .WK1 spreadsheet
format.
3.6 EpiInfo/EpiMap
A more sophisticated thematic mapping program can be downloaded from CDC (FTP)
which pairs EpiInfo (DB and Stat) and EpiMap (thematic mapping). I haven't had
time
to explore these, but hope to present them as well at ACSP. -- Jim Fisher
[log in to unmask]
wright.edu).
A cheap and simple solution for mapping developed by the WHO. It is completely
free
and easy to use, and self explaining. I use it myself for simple census data,
and I love it!.
The program can read Dbase data files or can be linked to a data manipulation
program
(EpiInfo), also by the WHO. You can download EpiMap from most SIMTEL-sites,
usually in the /pub/msdos/mapping directory or alike. -- Henk Vos
([log in to unmask])
3.7 Geo-EAS 1.2.1
Contact: Evan Englund
U.S. Environmental Protection Agency
Environmental Monitoring Systems Laboratory
Las Vegas, Nevada 89193-3478
According to Hsiu-Hua Liao ([log in to unmask]), EPA no longer distributes it.
He
offered to make a copy. It is PC base, and requires 3 HD disks.
Marion Shaw ([log in to unmask]) says, "personally, I would stay away from that
one
because it's extremely dated, methodologically. You may have difficulties
finding
software written for your applications and may have to translate concepts.
3.8 GEOEAZ (Is this Geo-EAS?)
I use GEOEAZ, an atrociously unfriendly program, for analyzing spatial
variograms and
for mixed model kriging which cannot be handled by SURFER. -- Julian Goulet
[log in to unmask]
3.9 GEO-PACK
A geostatistical package developed at the EPA's laboratory in Ada, Oklahoma,
includes
capabilities for disjunctive kriging and semi-automatic variogram modeling.
Contact: David M. Walters
U.S.EPA
Robert S. Kerr Environmental Research Laboratory
Ada, Oklahoma 74820
According to Hsiu-Hua Liao ([log in to unmask]) this is PC based and free.
You need
to send 2 pre-formatted, high density, 3.5" diskettes.
3.10 GeoQuery
Clinton Andrews ([log in to unmask]) likes this really user-friendly program
(Geoquery Corp, tel 708-357-0535) on the Mac platform. It does a variety of
spatial
analysis tricks (buffers, aggregations, etc.) but for a limited set of
applications. Its
primary target is sales managers for businesses, so planning applications have
to be
shoehorned in. But it's the easiest to use program he's come across that still
offers
analytical features. Students can be productive in less than an hour. He
marries it to a
wonderful stats program for exploratory data analysis called DataDesk (Data
Description,
tel 607-257-1000) using the publish-subscribe feature of the Mac OS.
3.11 GeoSim
From Ted Lyman, [log in to unmask]
You might want to look into software called GeoSim developed at Virginia Tech by
Departments of Computer Science and Geography. Several modules are available in
formats for DOS, Mac, Dec and Sparc. Try [log in to unmask]
3.12 GRASS
I worked a little with the versions of GRASS but of course they are very
demanding in
terms of hardware and would be difficult for students to master initially. --
Larry
Clowers, Department of Criminology, Sociology, Social Work, & Geography,
Arkansas
State University ([log in to unmask])
3.13 GS+
Jim Felley [log in to unmask] suggests this:
FROM: Gamma Design, Inc.
Box 201
457 East Bridge St.
Plainwell, MI 49080
Phone: (616) 685-9011
FAX: (616) 685-2345
Semivariograms, spatial autocorrelation, kriging Operating systems: DOS
3.14 GSLIB
A Unix package. Used at NC State, Raleigh.
3.15 IDRISI
3.15.1 Availability
IDRISI
Graduate School of Geography
Clark University
950 Main St.
Worcester, MA 01610
3.15.2 Description
Includes modules for autocorrelation analysis and trend surface analysis.
3.15.3 Comments
Does a lot of straightforward spatial statistics like autocorrelation and others
and is
entirely menu-driven. The tutorials can get you up and running fairly quickly.
For
simple things, it is a lot easier to learn than Arc\Info. It is fairly cheap
and can run on
most PCs in a DOS environment. The program does have its limitations, but it
might be
the best you can find. -- Alexander C Vias
I like this package and they give educational discounts. It's cheap (<$200, I
think) and
recently came out with a Windows version. -- Jim Felley <[log in to unmask]>
The handiest systems I am aware of are produced by people working in education.
IDRISI is an excellent PC based raster system, now available for a Windows
platform
which does a whole range of good things using remotely sensed images. From what
I
hear the latest version is jazzed up with all sorts of handy extras like a
"vectorize"
command and a pretty handy digitizing package. Better yet, the whole thing only
costs
about $300. No I don't get a commission on them. --
[log in to unmask]
3.16 Landview
Robbin R. Hough ([log in to unmask]) writes, I have found that most of the major
GIS
systems are beyond the bounds of single term courses in another subject, but I
am now
working with the version of LANDVIEW which comes with the newly released county
statistics package from the Bureau of the Census. It is a very well conceived
package and
may be the answer to courses like you are contemplating.
3.17 MapInfo
Fairly easy to learn but it also needs someone who is proficient with spatial
analysis and
mapping to do the support work and data preparation. -- Alicia L Allen
[log in to unmask]
Avoid. See comment under Atlas GIS, above. -- [log in to unmask]
3.18 MapViewer
The Dept of Geography at London Guildhall University uses this package. Etienne
BRESCH ([log in to unmask]) says, it "gives good enough results."
3.19 Minitab for Windows
The Dept of Geography at London Guildhall University uses this package. Etienne
BRESCH ([log in to unmask]) says the advantage is cheap licensing and
hardware.
Lyn Bloom also uses it for Spatial Data Analysis.
3.20 Pointstat
Net Levine has a program called 'pointstat' which takes x-y coordinate data and
constructs a number of indices of the spatial distribution (mean center,
standard distance
deviation, standard deviational ellipse, k-nearest neighbors, Moran's I, and
several distance
matrices - using spherical geometry). He's going to put this program out on the
Internet
in about a month or so. Contact Ned at [log in to unmask]
3.21 PMAP
I sometimes use PMAP for just the kind of thing you want to do, that is map
algebra on
maps whose cells are categories. (Shortest path downhill, cost based distance,
etc etc).
It has poor presentation results, though can handle color patches on screen. I
have written
SPSS macros to move data between these several packages (the others are SURFER,
GEOEAZ, and SPSS for Windows -- MF).
PMAP comes in a student version. Its data could be passed in and out of
SAS.
Character symbols on a printer are its easiest output. However, it does no
spatial
statistics. It does do spatially weighted cross tabulation stuff. -- Julian
Goulet,
[log in to unmask]
3.22 PVWave
Karsten Self ([log in to unmask]) suggests PVWave - data visualization
software for
Unix workstations - as a potential source of info and a white-hot demo package
(you
supply the workstation). PVWave is produced by someone somewhere in Colorado.
Can't get more specific than that -- you might find them by asking SAS-L.
3.23 R Programs
3.23.1 Description
Does univariate spatial autocorrelograms; simple and partial Mantel tests;
multivariate
correlograms; clustering with space and time contiguity constraint; connection
networks;
contingency periodogram, chronological clustering; spatial autocorrelation,
geographically
constrained ANOVA.
3.23.2 Platforms
Currently runs on MacIntosh, IBM mainframe, MacIntosh, VAX. According to Pierre
Legendre, its author, a Windows version should be available by the end of the
summer,
with a nice user's interface.
3.23.3 Obtaining
Pierre Legendre
Departement de Sciences biologiques
Universite de Montreal
C.P. 6128, Succursale A
Montreal, Quebec, Canada
H3C 3J7
E-mail: 3771 at UMTLVT
FAX: (514) 343-2293
3.23.4 Comments
It's very friendly to use and has folders and routines designed for teaching.
Legendre has
been teaching a course on spatial analysis for undergraduate students for
several years. -
- Miquel Angel Rodriguez, Universitat de Barcelona, [log in to unmask]
3.24 REGARD (from Antony Unwin -- [log in to unmask])
Together with Graham Wills and John Haslett I developed the REGARD software
package
in Dublin and work on it has moved with me to Augsburg. REGARD is specifically
designed for exploring spatial data sets with interactive graphic tools and is
particularly
effective for collaborating with users. I have run short courses using REGARD
which
have worked well because we encouraged people to bring their own data sets.
Since
REGARD only runs on Macs and is research software (i.e. we have a nice manual
and
the package is robust but it is not shrink-wrapped) this may not help you a lot,
but I shall
be presenting a paper at the Interfaces conference in Pittsburgh including a
live
demonstration of applying the package.
3.25 Regional Analysis System (RAS)
The program we use in our class is RAS (Regional Analysis System) developed by
ourselves, which is integrated with ARC/INFO and has the capabilities of
defining spatial
weight matrices, calculating Moran I, G and LISA and graphic analysis. It is on
the Sun
Station instead of PC. We have a paper on RAS and some applications of spatial
statistics Bao (1995) in which you may find some reference about Getis and
Anselin's
papers. (From Shuming Bao, [log in to unmask]).
3.26 S+
See Art Getis' comment above. S+ is flexible and can do maps. It does help to
have
some understanding of mathematics used in statistics since its data structures
correspond
to them quite closely.
I agree with you that S+ is too difficult (or too intimidating) for many (incl.
me). --
Thurman Wenzl, [log in to unmask]
3.27 SAS
3.27.1 Pro
Andy Eppard Barnett ([log in to unmask]) writes:
From a 'student' viewpoint, I would recommend that a sufficiently powerful
package (like
SAS) be used, especially one that is used widely enough to help in obtaining a
job. It is
true that too much time could be spent on technical details like how to produce
a map,
but could this not be delegated to a syllabus, appendix, or user manual that
even the
students themselves added to? Just a series of 'canned procs' to do certain
tasks, like
producing a map.
There is plenty of documentation by/for SAS on this kind of thing already,
so much
that finding it is more the problem than not. If anything, I think the biggest
factor with
SAS is cost, but universities get a much reduced cost for a license than
commercial users.
Finally, if computer science itself presents a technical problem, there are
a number of
good introductory texts on computers, O/S's, etc. that might help. It has been a
long time
since I was a student, but a book that I picked up recently at a used store that
I like is
"System Architecture: Software and Hardware Concepts", by Leigh and Ali, 1988.
It
would be a good, general, relatively complete reference for non-computer science
students
to use as an intro to computing. There are plenty of diagrams, and the text is
very easy
to read. Perhaps it would be good to list as a reference in a syllabus, or to
place in a
reserve room somewhere.
3.27.2 Con
David Fosdecki writes, "we struggled through sas with weekly project exercises.
I
abandoned it midway through the semester and used SPSS for windows."
3.28 SPACESTAT
3.28.1 Overview
A software program for the analysis of spatial data, developed by Luc Anselin is
available
from the Regional Research Institute at West Virginia University. The current
version is
1.50 (latest revision 9/8/94).
3.28.2 Current Capabilities
- data manipulation utilities (moving data from ascii files into the format
used by
GAUSS - SpaceStat is compiled in GAUSS)
- spatial weights utilities: the construction, standardization, manipulation
of spatial
weights matrices, for irregular and regular lattices, computations of roots
and other
characteristics of the weights matrix; distance-based weights matrices
(including great
circle distance)
- descriptive statistics: mean, variance, standard deviation, quartiles,
interquartile
distance, third and fourth moment, test for normality, correlation
coefficients
- join count statistics for spatial autocorrelation
- Moran's I for spatial autocorrelation: normal, randomization and permutation
distributions, correlogram, Moran scatterplot, local Moran (Moran LISA)
- Geary's c for spatial autocorrelation: normal, randomization and permutation
distributions, correlogram
- Getis-Ord G and Gi statistics for local spatial autocorrelation
- Hubert-Golledge QAP statistics (implemented for generic case, as well as for
Moran,
Geary and absolute differences)
- Wartenberg multivariate spatial autocorrelation
- Least squares regression with diagnostics for spatial dependence (including
Moran's
I, Lagrange Multiplier tests, Kelejian-Robinson test)
- robust least squares regression (Jackknife)
- maximum likelihood estimation of regression models with spatial
autoregressive errors
and spatial autoregressive dependent variable (with diagnostics)
- GMM estimation of models with spatial autoregressive dependent variable
- bootstrap estimation of models with spatial autoregressive dependent
variable
- two-step and ML estimation of heteroskedastic models (groupwise
heteroskedasticity,
random coefficients)
3.28.3 System Requirements
The current version of SpaceStat is written in GAUSS and is distributed with a
GAUSS
Runtime Module (version of GAUSS currently used is 3.2.4). It requires a 386 or
better
PC with math coprocessor and at least 4 megs of RAM (the more the better).
(GAUSS is a product of Aptech Systems, Inc. of Kent, Washington)
3.:\U Updates
Version 2.0 of SpaceStat is expected to be ready in summer 1995. Current users
will be
able to obtain a low-cost upgrade. The new version will include the capacity to
handle
large data sets (several thousands of observations) for most descriptive
statistics and
diagnostics (not ML estimation).
3.28.5 Pricing
(NOTE: SpaceStat is no longer distributed by NCGIA Santa Barbara, prices are
effective
August 1, 1994, and subject to change without notice - prices advertized in the
NCGIA
Brochure will be honored until September 1, 1994, for all orders forwarded by
NCGIA)
Current prices are for PC version only.
Single user (includes Gauss Runtime module, manual/tutorial and sample data
sets):
- academic: $250
- government: $300
- private sector: $500
Multiple Licenses:
- academic: site license $1,500 - includes one master copy of SpaceStat,
master copy
of manual and two bound copies and Gauss Runtime modules (one
disk for
each installed copy - specify number needed)
- others: contact for information
Tutorial only: $45 per copy
Update to Version 1.50, for registered users: $25 (includes Revision Notes and
new Gauss
Runtime Module - prepaid orders only, the original SpaceStat disk must be
returned with
order)
All prices include U.S. Post surface mailing. Non U.S. orders, add $10 for
airmail. Ask
about express rates.
All payment in U.S. dollars only - pre-payment preferred and required for non-US
orders.
For ordering and other information contact:
Luc Anselin
Regional Research Institute
West Virginia University
511 North High Street
P.O. Box 6825
Morgantown, WV 26506-6825
e-mail: [log in to unmask]
tel: (304) 293-8546 or 293-2896
fax: (304) 293-6699
3.28.6 Comments and Use
Spacestat might not be user-friendly enough for the proposed class, but it is
powerful.
-- Marlon Boarnet, UC-Irvine, [log in to unmask]
A very good spatial statistics program that does extremely sophisticated
analysis and can
be linked up to IDRISI or ARC\Info. However, it may be too advanced for what
you
want. -- Alexander C Vias
One of the best programs. Calculates various statistics based on areal or zonal
distributions. The program can calculate measures of spatial autocorrelation
(Moran' I,
Geary's C), local indicators of spatial association (LISA), 'classical'
multivariate
regression, trend surface analysis, and spatial lag regression. The latter is
one of
Anselin's theoretical developments and allows the incorporation of a spatial
proximity
variable (which he calls a 'spatial lag') into a regression model; he estimates
the model
using maximum likelihoods. The program has its idiosyncracies, but is very, very
powerful. -- Ned Levine.
3.29 SPANS
FROM: TYDAC Technologies Corp.
1655 N. Ft. Myer Dr., 3rd Floor
Arlington, VA 22209
Includes a language allowing user to create analytical routines. Some spatial
analysis
implicit in modules (e.g., some modules return a value of Moran's I).
Expensive!! -- Jim
Felley <[log in to unmask]>
3.30 SPSS for Windows and Add-Ons
David Fosdecki suggests SPSS for Windows ($149, student ver.) with Mapinfo($259,
educ.) and the map procedure (a $50, mapbasic app to connect spss/mapinfo)
option along
with Norusis. The map procedure adds a menu option allowing SPSS to send data
to
mapinfo and vice versa in mapinfo. Data sent to mapinfo become a table (mapinfo
slang
for unmapped layer) which then can be viewed through thematic map. I've used it
in
geodemographic or lifestyle cluster and factor analysis. Regression residual
spatial
mapping. It's slick and the system is affordable to students.
<[log in to unmask]>
3.31 STAC
Ned Levine says the Illinois Criminal Justice Information Authority has produced
a simple
program called STAC (spatial & temporal analysis of crime) which they distribute
to
police departments throughout the country. The program calculates mean centers,
standard deviational ellipses, and 'hot spots' given a set of x-y coordinates.
3.32 Statview
"A nice, easy stats program for the mac." It makes graphs and charts easily.
There may
be a IBM version.
3.33 SURFER for Windows (Golden Software).
An outstanding gridding and display package. However, you are more interested,
I
presume, in choropleth (sp?) maps. [Not really, I'd like graduated point symbol
maps
much more). SURFER does not handle those very well. It can, but it is not as
user
friendly as contour maps. -- Julian Goulet ([log in to unmask]).
A likely candidate on the PC for mapping. -- Pierre Legendre
([log in to unmask]
.ca)
3.34 SYSTAT
I am fairly convinced that SYSTAT offers the best value in classroom software.
It's
available for Windows but there is still (I think) a DOS version. The Windows
version
is menu driven (there's actually an internal version of BASIC for those who want
to
incorporate non-standard operations. SYSTAT is now owned by SPSS but is cheaper
and
just as powerful. It is more analysis oriented than SPSS, that is SPSS offers
more in the
way of data presentation, SYSTAT is good for data analysis. It's a bit of a
toss-up
though as SPSS is probably more common in the work place. Bruce Wm. Pigozzi
([log in to unmask])
John Cook ([log in to unmask]) says SYSTAT had a very inexpensive intro
version
called MYSTAT which he used to get free at Carolina (may have been part of a
site
license). MYSTAT was very basic descriptive procs, some nonparametric
correlation
things, and basic regression. The editor was not the greatest, but adequate.
3.35 USTAT - Jerry Schneider ([log in to unmask])
A DOS program that calculates about 25 spatial statistics and also produces a
set of charts
and maps that can be used to see those measures that can be plotted. Requires
data, by
small area and time point, and a boundary file as input. It is about as simple
as it could
be. The hard part is getting the data cleaned up (i.e. adjusting for zonal
changes over
time, etc.) and interpreting the results. I use it to measure spatial change
over time
(magnitude and shape). The charts/maps are produced as a DXF file which then
can be
pulled up in AutoCAD for editing and plotting, in color or B/W. I have been
using it in
the classroom for some time now.
3.36 Other
3.36.1 John Cook ([log in to unmask]) thinks Lotus also recently
released some
kind of basic mapping package, and Harvard Graphics had their Harvard
Geographics for
a while.
3.36.2 Robbin R. Hough ([log in to unmask]) suggests using Microsoft
Office
suite especially Excell and Access as the basic tools for data exploration.
3.36.3 Works for windows provide a series of
wordprocessing/spreadsheets/graphics sub-packages that are very user-friendly
and
'student-proof', ie difficult to mess up! For stats and analysis, we use
Minitab for
Windows and Excel. The advantage of those packages is that the license works out
quite
cheap, but hardware or machines don't and in the end it takes a concerted effort
within
the department to want to develop something the students will benefit from
(those famous
transferable skills). -- [log in to unmask]
3.36.4 Geoff Selig ([log in to unmask] or [log in to unmask])
said
there was a mention of a WWW site that had several programs that did exactly
what you
want in the edstat list a while back. I will try to track down the URL for the
site.
3.36.5 [log in to unmask]
I would recommend a system produced by the Geography Department at Sheffield
University (I think). It walks the user through some carefully contrived
exercises in
ArcInfo and IDRISI. However, none of ESRI's killer command structures are
required,
so its a lot easier to work with. The contact I have for that is
[log in to unmask]
AC.UK (who is really Ms. J. Burgan, tel 0742-725-248.
4 Miscellaneous Comments
4.1 On the Utility of Such A Course and How to Package It
4.1.1 [log in to unmask]
There are several points worthy of comment in your posting. First and foremost,
you need
to talk to some geographers! All courses that involve a significant software
component
require some computer skills. Unfortunately the majority of spatial analytical
howlers
result from a failure to understand either a) the nature of spatial data or b)
the techniques
used to process those data and produce graphic images from them.
4.1.2 Thurman Wenzl, Natl Inst for Occ Saf and Heal,
[log in to unmask]
I think there is wide interest in what you come up with; do post a summary.
I've seen
an occasional ref in environ monitoring stats which might apply to your
question/plan, but
I can't think of the author right now;
Perhaps the only solution is a 2 step process: a first course with only
concepts, setting
up the problems, and interpretation of outputs; then (in a follow-up course),
some
discussion and use of software and how to do the analyses - for a subset of the
students.
4.1.3 Etienne Bresch, Dept of Geography, London Guildhall University,
[log in to unmask]
A couple of years ago I had a similar problem, ie about the design of a new
series of
courses to be planned as a three-year cycle, or basically following our degree
progression.
We do teach some of the material you are mentioning in your message, but it
is more
directly related to cartography. One thing I have learned is that students seem
to be more
confident with the computers only once they know exactly what it is they are
supposed
to be doing. Just to give you an example, if they can work out what a standard
deviation
is and how and why you calculate it (yes, it takes us back to basics, with all
those
students with varied backgrounds), using a simple set of population density data
of a small
town so they can calculate everything by hand using a pocket calculator, then
they waste
much less time when it comes to do a similar exercise by computer.
4.1.4 Rich Winterfield ([log in to unmask])
I think that your proposal can provide a valuable resource for your students. I
am a GIS
Analyst working for the University of Tennessee, Knoxville at the Oak Ridge
National
Laboratory. I received my MSP from UT, Knoxville in 1992 and went right into
GIS
work. I work primarily with Arc/Info.
My initial thoughts are that over the long run, you might want fold a
majority of this
course into the Intro to Stats and Research Methods courses. Then you'd be able
to have
students use the skills in a capstone methods course. In the short term, I
would not be
so quick to dismiss some of the more sophisticated software products. While you
are
correct that they would have to be user friendly to avoid making the software
the course,
these products can all be customized to present only what you want them to. I
believe
that you could get some cheap/free assistance from some of the software
companies that
advertise their commitment to education, and from users in the world.
Having personally experienced the problems associated with a weak
statistical
background when dealing with spatial data, I would consider offering my
assistance in
some programming in Arc/Info or ArcView to help customize these products to
match
your lesson plans. If you decide to go this route, and are looking for help,
drop me a
line, and we'll see if I can be of assistance. If not, keep going with your
plans, the
students need this information.
4.1.5 Pat Cabe, Dept of Psych, Pembroke State University
([log in to unmask]
embroke.edu)
This is a good idea, there are apparently people "out there" who are doing
it/teaching it,
it seems to have lots of applications (even outside the vaguely geographical
realm you
might have in mind), and I hope you'll seriously pursue it.
4.1.6 Paul Cote ([log in to unmask])
The course you describe sounds very worthwhile. And I don't have any personal
experience of a course like this, but I would be interested in what you turn up,
especially
re. textbooks. I have a feeling there maybe some promising work going on at
Rutgers,
New Brunswick where Lyna Wiggins and Richard Brail are both working.
4.1.7 John Cook ([log in to unmask])
Your class sounds like it could be fun, and very useful for students.
4.1.8 Gary M. Greenberg ([log in to unmask])
There is a pretty brilliant guy in UF's geography dept. named Tim Fik who might
offer
great advice on making quant. methods really interesting to learn. He may be a
useful
contact as he had some Public Domain quant. tool that he bragged on <though I
graduated
before I could take the course which he was to use it for; maybe I'll do that
for fun).
Finally, make 'em work! Scare off the babies; grad school is no place to
coddle the
minds and hearts of tomorrow's workforce (but then I suspect you know that
having
challenged my thinking fairly rigorously one summer over the wire -- thanks for
the
backhanded complement, Gary. MF).
4.1.9 Antony Unwin ([log in to unmask])
Good luck! The aims of your course seem both sensible and practical, how easy
they will
be to achieve is another matter. I would like to hear how you get on.
I have been interested in exploratory spatial data analysis (which could be
an
important part of your course) for a while because it seems to have real
potential for
application in comparison to many of the more mathematical approaches. Thus I
find the
book by Cressie valuable for telling me what progress has been made
theoretically but I
cannot imagine many of the models discussed bringing much insight for real data.
You also mentioned spatial autocorrelation. The main reservation that I
have there
is that it is often assumed that it is the same across a whole data set. Local
statistics are
worth looking at (e.g. recent work by Getis and Ord or by Luc Anselin) but I
always find
the interactive graphical methods more powerful anyway because autocorrelation
can only
measure a very limiting type of spatial association and there can be so many
different
types.
Some packages produce nice maps and claim to do statistics but as you will
have
gathered from the above I am not keen on static presentation graphics. Too
often the
beautifully drawn chloropleth maps turn out to be heavily dependent on the
classification
used and even the cleverest tricks of the cartographers can not reproduce the
multivariate
structures that are important (another hobby horse of mine: spatial data are
inherently
multivariate and a mixture of different data types recorded in irregular
patterns, theoretical
models posit single continuous variables recorded on a regular grid). In
principle you
could draw several maps and choose the best (which is what you in effect do with
interactive graphics fast and flexibly) but the effort and the time make this
impractical.
Antony goes on to discuss his REGARD software.
4.1.9.1 My Comment
I'm sorry I mentioned spatial autocorrelation. I usually introduce it in
low-level courses
as a warning how a little knowledge can be worse than no knowledge at all.
However,
in terms of the course I have in mind, I'm much more in agreement with Antony.
Intuitive, interactive, exploratory tools seem much more appropriate for MCP
students.
4.2 On How The Course Fits into the Planning Curriculum
4.2.1 Robert Lincoln, Michigan State University ([log in to unmask])
Glad to see that someone's taking this on directly. Here at MSU, we're talking
about
having a spatial analysis course or part thereof in our methods sequence and
we're
running into the same issues: noone's really done this for planning, so most of
the course
material out there is specialized for geographers, etc. In particular, noone
seems to have
put together data and analysis "kits" for the kinds of functions most useful to
planners
(though it may be that the new "First St." product that combines TIGER, ArcView
2 and
a bunch of jurisdiction, tract, and block census variables will fill some of the
bill.
One thing you might consider: as I understand it, ESRI is making ArcView 1
freely
available by FTP. Maybe the thing to do is to build the course around some
standard
maps and data sets that could be generated pretty easily with ArcInfo for PCs
and the
TIGER CDs, and then "teach" with the mapping software and the related flat-file
databases.
I wonder, though, what the real focus should be? Should we be teaching our
students
to map data better, e.g. getting them in the habit of geocoding everything and
letting the
software pretty much do the rest? Should we be spending a lot of time asking
whether
geographic variables like proximity should be analyzed along with other
statistics (are
urban and regional planners to do analysis of future land values around
landfills based on
historic trends?)? Should we be simply teaching them to think about physical
and social
phenomena more thoroughly by presenting and analyzing them on a geographical
basis?
I go back and forth on a lot of these questions, seeing long term possibilities
for making
planners more useful but wondering what skills are really useful now.
So, I hope I've added one piece of useful information and a good question or
two.
Where are you going with your course? And, the ubiquitous question, what are
other
people out there doing? How about some working planners -- what skills (both
conceptual and task oriented) do you wish that you had, or the planners working
for you
had, in this area?
4.2.2 Abhijit Bhahmachari ([log in to unmask])
It sounds like a real nice course, and I think, if offered, the course will
surely help the
planning students. But, for some reason, I thought that it might become a
little too much
for a semester long course. Instead, if you find a real good, user-friendly
(very imp.)
software, learning the software may be a course in itself. Once the students be
come
familiar with the software (mid semester) they may be asked to pick one of the
topics
from i) social, ii) environmental, iii) demography, etc., and do a semester
project using
the software, and available data from the selected topic. This will probably
reduce data
collection time. Since you mentioned that the course is being designed for a
SECOND
year (I did not, I said second-semester -- MF) student, I think, you may not
want to spend
more time on accessing data. Even if the data is supplied, decoding, analysis,
and
presentation of the supplied data using the software would be a nice course.
-- Abhijit is a student here in our department (MF).
4.2.3 From an Anonymous Staff Member in the Santa Barbara Planning
Department (thanks to Barbara Sommer, [log in to unmask])
There used to be several course at UC Santa Barbara in the Geography Department.
The
course should be "not too theoretical. Need practical, real world data
sets/issues where
students can enjoy thinking about viable hypotheses, not just manipulating data
which are
uninspiring.
4.3 Other
4.3.1 Rick Dalzell ([log in to unmask])
You should see all the stuff that the local high school teams are doing in the
EarthVision
Project on the EPA's Cray computer. Believe similar projects are going on
around the
country. Contact Mark Neal, Central High School, Bay City, MI 48708
(517-893-9541)
or Dr. Ralph Coppola, Special Projects Director, S.V.S.U., University Center, MI
48710.
4.3.2 Larry Clowers ([log in to unmask])
I've been thinking along avenues similar to your spatial analysis course. I am
hopeful of
(using) ARC/View on a fast pentium. It seems to be easily learned and a lot of
data is
available in appropriate formats. I would like to know how your plans develop.
4.3.3 Karsten Self ([log in to unmask])
Working for industry and trying to dream up ideas for visualization, I'd be
interested in
receiving a summary of what you find.
4.3.4 Bill Westbrook ([log in to unmask])
I suggest you contact NASA at their Goddard Space Flight Center for textual
materials
and possibly the needed software. May I assume you are referring to Remote
Sensing of
Earth Resources? If so, you might contact the Office for Remote Sensing of
Earth
Resources at Pennsylvania State University.
4.3.4.1 My Comment
No, I'm not at all interested in remote sensing of earth resources.
4.3.5 SPACE-L
Marlon Boarnet ([log in to unmask]) suggests posting the question to the SPACE-L
listserver (he thought that's the name).
5 Results of a WebCrawler Search
Melvin Klassen ([log in to unmask]) was kind enough to forward the following
results of a WebCrawler Search.
<p>The query "spatial data analysis course" found 87 documents and returned 25:
<p><tt>1000
</tt><a href="http://daac.gsfc.nasa.gov/DATASET_DOCS/dao_dataset.html"
>DATA ASSIMILATION OFFICE 4D TIMESERIES MULTI-YEAR ANALY-
SIS</a><br>
<tt>0968
</tt><a href="http://eosdis.larc.nasa.gov:12000/dataset_documents/srb.html"
>http://eosdis.larc.nasa.gov:12000/dataset_documents/srb.html</a><br>
<tt>0829
</tt><a href="http://daac.gsfc.nasa.gov/DATASET_DOCS/tovs_dataset.html"
>TOVS Pathfinder Path A Guide</a><br>
<tt>0743
</tt><a href="http://www.blm.gov/gis/nsdi.html"
>NSDI MetaData and WWW Mapping Sites</a><br>
<tt>0735
</tt><a href="http://www.essex.ac.uk/social-science-methodology-school"
>Social Science Methodology School Home Page</a><br>
<tt>0681
</tt><a href="http://www.geo.ed.ac.uk/home/giswww.html"
>GIS WWW Resource List</a><br>
<tt>0658
</tt><a href="http://www.esd.ornl.gov/esd-research-org.html"
>Environmental Sciences Division Research Organization</a><br>
<tt>0588
</tt><a href="ftp://ftp.csn.org/COGS/ores.txt"
>ftp://ftp.csn.org/COGS/ores.txt</a><br>
<tt>0585
</tt><a href="http://rigel.csuchico.edu/ores.html"
>On-line Resources for Earth Scientists (ORES)</a><br>
<tt>0554
</tt><a href="http://www.cast.uark.edu/d.cast/cast.html"
>CAST Home Page</a><br>
<tt>0541
</tt><a href="http://haleakala.jpl.nasa.gov/atmo.html"
>ATBD for Surface Radiance and Reflectance</a><br>
<tt>0533
</tt><a href="ftp://ftp.nwnet.net/user-docs/government/gumprecht-guide.txt"
>ftp://ftp.nwnet.net/user-docs/government/gumprecht-guide.txt</a><br>
<tt>0528
</tt><a href="ftp://geology.wisc.edu/pub/inqua/newltr-3.txt"
>ftp://geology.wisc.edu/pub/inqua/newltr-3.txt</a><br>
<tt>0487
</tt><a href="http://www.geog.nott.ac.uk/remote/faq-2.html"
>http://www.geog.nott.ac.uk/remote/faq-2.html</a><br>
<tt>0455
</tt><a href="ftp://fgdc.er.usgs.gov/gdc/Forum/Geodata.pgm.txt"
>ftp://fgdc.er.usgs.gov/gdc/Forum/Geodata.pgm.txt</a><br>
<tt>0440
</tt><a href="ftp://geology.wisc.edu/pub/inqua/newltr-5.txt"
>ftp://geology.wisc.edu/pub/inqua/newltr-5.txt</a><br>
<tt>0430
</tt><a href="http://newsroom.hitc.com/tmdb/Prot_did318.html"
>Status Report</a><br>
<tt>0424
</tt><a href="ftp://umd5.umd.edu/pub/vrtp/Jove/Articles/dsgbjsbb.txt"
>ftp://umd5.umd.edu/pub/vrtp/Jove/Articles/dsgbjsbb.txt</a><br>
<tt>0398
</tt><a href="http://lib.stat.cmu.edu/S"
>StatLib---Software and extensions for the S (Splus) language</a><br>
<tt>0380
</tt><a href="ftp://abraxas.adelphi.edu/gis/FAQ"
>ftp://abraxas.adelphi.edu/gis/FAQ</a><br>
<tt>0370
</tt><a href="http://www.ncgia.ucsb.edu/core_curriculum.html"
>The NCGIA Core Curriculum</a><br>
<tt>0357
</tt><a href="http://ncgia.umesve.maine.edu/tech.html"
>NCGIA Technical Report List</a><br>
<tt>0341
</tt><a href="http://cirrus.larc.nasa.gov/fire/ASDfire.html"
>The First ISCCP Regional Experiment</a><br>
<tt>0287
</tt><a href="http://www.ckm.ucsf.edu/Papers/SIGBIO-93"
>Processing Cross-Sectional Image Data for Reconstruction of Human
Developmental Anatomy from Museum Specimens.</a><br>
<tt>0284
</tt><a href="gopher://gopher.towson.edu/00/
Guides%20to%20Information%20Resources/
Frequently%20Asked%20Question%20Lists%20%28FAQs%29/
File%20Compression%20part%202"
>gopher://gopher.towson.edu/00/Guides%20to%20
Information%20Resources/Frequently%20Asked%20Question%20Lists%20%28FAQs%29/
File%20Compression%20part%202</a><br>
<hr><p>
6 Texts on the Subject and References
6.1 General Comments
6.1.1 Ned Levine
There is a huge gap between very elementary texts and mainstream spatial
statistics texts;
there is nothing intermediate which would be appropriate for planning and
geography
graduate students. At the simple level, the books by Ebdon and Hammond &
McCullogh
cover many elements of spatial statistics (e.g., Nearest Neighbor indices,
descriptive
indices, Moran's "I"). The next level, however, is very technical and requires
a good
math background. I would suggest books by Haining (1990), Cressie (1991) and
Venables & Ripley which cover much of the modern developments in spatial
statistics.
Ripley (1980), while a classic, is very, very difficult to follow without a
solid
math/statistics background.
6.1.2 Jon Claerbout ([log in to unmask])
In Geophysics all our correlations are spatial (and temporal). Be assured that
there are
many specialized techniques. Please enjoy the free books below.
Jon Claerbout: FREE BOOKS http://sepwww.stanford.edu/sep/prof
6.1.3 Etienne Bresch, [log in to unmask]
I have noticed that a lot of students need to start from scratch, while others
cruise more
quickly.
6.1.4 Terry Gilhula, Dept. of Geog, Univ. of Tennessee and Metropolitan
Planning Commission ([log in to unmask])
I have a brief recommendation for a cost-effective, simple computer software
that is most
appropriate for planning students, allowing, among numerous topics, population
projections, economic base analysis, real estate investment analysis, and many
other
applications. The source of this "stuff" is Klosterman, Brail, and Bossard
(1993). This
includes a text and accompanying Lotus 1-2-3 based spreadsheets that are
pre-programmed
for the above mentioned applications. An accompanying text is Klosterman (1990)
which
provides the rationale behind the techniques in the spreadsheets.
I have used the materials in school as a grad student, and professionally as
a practicing
planner. Perhaps Klosterman, et al should send me a royalty for plugging their
work, but
it really is pretty straightforward and useful material.
I saw a catalog recently with this material advertised. Sells for about
$40, software
included.
6.2 References
Anselin, Luc. c1988. Spatial econometrics: methods and models. Boston: Kluwer
Academic Publishers.
Bao, Shuming. 1995. Paper on RAS and applications of spatial statistics.
Computer,
Environment, and Urban Systems 19(1).
Bampton. The Poet's Guide to ArcInfo.
"A brilliant, humorous and insightful book that I am currently writing
myself. --
[log in to unmask]
Batty, Michael and Paul Longley. Fractal cities.
Bendavid-Val, Avram. Regional and Local Economic Analysis for Practitioners.
Borcard & Legendre. 1994. Article in Environmental and Ecological Statistics, n
1.
Clark, William and Hosking. Statistical Methods for Geographers. Wiley.
-- Clark is at UCLA. Megan Blake says it makes little reference to
spatial
autocorrelation. Alexander C Vias has heard good things about it.
Cressie, Noel. c1991. Statistics for spatial data. New York: J. Wiley. ISBN
0-471-84336-
-9 $90.00 (!!!!!)
Chrisman, Nick. c1996. Geographic Information Analysis.
Written to give an understanding of different 'GISs' in context - starts out
with
measurement and representation. Explains concepts of what operations and
transformations are done in different GISs - (quite a bit of time on this -
what goes
on in GIS) - then last chapter discusses social and institutional context.
(most
examples given are in some context - very few 'black box' illustrations). I
liked the
strong conceptual organization (I'm a sucker for that sort of thing) - from
measure-
ment to social/institutional context. Nick's e-mail address is
[log in to unmask]
edu -- Tara Clapp ([log in to unmask]).
Davis, John C. 1986. Statistics and data analysis in geology, 2nd ed. John
Wiley.
Deutsch, Clayton V. and Andre Journel. 1992. Geostatistical software library and
user's
guide". Oxford U. Press.
-- Comes with Fortran code.
Ebdon, David.
ESRI. The ARC Macro Language Workbook.
. Understanding GIS.
Earickson and Harlin. 1994. Geographic measurement and quantitative analysis.
Macmillan.
-- I've been most satisfied with Earickson and Harlin for the basic
course (there's
also a multivariate oriented second course) in a semester format
(compared to
McGrew and Monroe or Clark and Hosking). -- Bruce Wm. Pigozzi
Frank, Camparai and Formentini, ed. 1992. Theories and methods of
spatio-temporal
reasoning in geographic space. Springer Verlag.
Garson and Biggs. Analytic Mapping and geographic databases. Sage.
Golledge Reginald G. and John N. Rayner 1982. Proximity and Preference: Problems
in
the multidimensional analysis of large data sets. U. of Minnesota Press,
Minneapolis.
Golledge is at UC Santa Barbara and Rayner is at Ohio State. You might
try to
reach one of them. -- Harry Kitchen ([log in to unmask])
Griffith, Daniel A. and Carl G. Amrhein. n.d. Statistical analysis for
geographers, with
contributions by Joseph R. Desloges. Englewood Cliffs, N.J.
Hammond and McCullogh.
Haining, Robert. 1990. Spatial data analysis in the social and environmental
sciences.
Cambridge: Cambridge University Press.
-- More elementary than Cressie. Most of the examples come from the social
sciences
and economics (with some environmental data too). -- Dale Zimmerman
(dzimmer@s-
tat.uiowa.edu)
Hanushek and Jackson. Statistical Methods for Social Scientists.
-- I've used this book myself, and I can attest that it's excellent.
However I
would not use it as a first course in statistics, and for a second only
with better
students. Little discussion of spatial statistics or data analysis.
Huxhold, William.
Isaaks and Srivastava. An Introduction to Applied Geostatistics.
Johnston, R.J. 1980. Multivariate statistical analysis in geography. Longman
Scientific
and Technical.
Klosterman, Richard. 1990. Community Analysis and Planning Techniques.
Klosterman, Richard, Richard K. Brail, and Earl Bossard. 1993. Spreadsheet
Models for
Urban and Regional Analysis.
Lapointe & Legendre. 1994. Article in Applied Statistics, 43:237-257
McGrew, J. Chapman and Charles B. Monroe. 1993. An Introduction to Statistical
Problem Solving in Geography. Wm. C. Brown Publishers.
"The book has been a great resource as I design my analytical methods
for my
thesis" -- Lynn Rosentrater.
Martin, David. Geographic Information Systems and their Socioeconomic
Applications.
"An excellent text" -- [log in to unmask]
Matthews, H. and Foster, I. 1989. Geographical data sources, analysis and
presentation.
Oxford University Press.
"It's fairly basic, but there are a lot of interesting exercises you can
use directly.
It has the huge advantage of being very cheap (about British # 8.50,
which works
out probably at US $ 2.00)." -- Etienne Bresch, [log in to unmask]
Monmonier. Mark. How to Lie With Maps.
"Has an interesting angle on how you can (purposely or innocently) skew
data
when performing spatial analysis." -- Randy Fusaro
"The classic text" -- [log in to unmask]
Myers, Dowell. Analysis with Local Census Data
Norusis, Marija J. 1990. The SPSS Guide to Data Analysis. SPSS Inc.
John Odland. Spatial autocorrelation. Sage.
-- "A very nice treatment". Barbara Warner, [log in to unmask]
Plane and Rogerson. The Geographical Analysis of Population, With Applications
to
Planning and Business.
"A very useful text for teaching spatial data analysis ... from a
planning and
demographic perspective" -- Bruce Newbold, Geography, U. of Illinois.
Ripley, 1980. Spatial statistics.
Sage Publications series on Scientific Geography.
"It is comprised of about 8 little books on issues such as
transportation, population
and urban spatial structure. Some of the books are more complex than
others."
-- Jim Fisher, ([log in to unmask])
"You might want to consider a couple of Sage books as supplements. See
Garson
and Biggs and Odland. -- Barbara Warner, Sociology, University of
Kentucky
([log in to unmask])
Scholten, Henk J and Stillwell, John C.H. Geographical Information Systems For
Urban
And Regional Planning Kluwer Acad. Press.
-- Very good reading to link GIS conceptually to modelling with
quantitative data
(Gary M. Greenberg ([log in to unmask])
Upton, G. and Fingleton, B. 1985. Spatial Data Analysis: By Example. Chichester:
John
Wiley.
-- Most of the examples in this book come from biology/ecology (Dale
Zimmerman, [log in to unmask])
Venerables and Ripley. Modern applied statistics for S-Plus.
-- Ned Levine says, it has a chapter on spatial statistics which
concentrates on both
contour estimation and interpolation of three-dimensional surfaces.
They have
written some S-Plus routines which calculate these functions which, I
believe, are
in statlib.
Walford, N. 1995. Geographical Data Analysis. J.Wiley & Sons.
-- Covers very much the material you want. -- Etienne Bresch,
[log in to unmask]
.ac.uk
Marsh Feldman Phone: 401/792-5953
Community Planning, 204 Rodman Hall FAX: 401/792-4395
The University of Rhode Island Internet: [log in to unmask]
Kingston, RI 02881-0815
"Marginality confers legitimacy on one's contrariness."
|
