statistical packages, gis and hypotheses

[John Gary]

How ya do?  I'm working on a Master of Urban Planning @ KU. My professors speak with certitude that SPSS and ArcGIS are the only options when it comes to statistical and geographic analysis. The Linux/Gnu alternative appears to be GRASS GIS and R statistics. Does anyone have experience using these software packages? 

As a side issue, any tips on crafting a hypothesis? For my purposes, I have to use data from the census or similar source. The question has to be proven, or not-(I'm confused about the null hypothesis) using census data. I had previously thought of what I consider to be a great question. Does the greater occurrence of liquor stores in minority neighborhoods lead to increased consumption of alcohol? Prima facie this phenomena is the result of zoning. More affluent neighborhoods will likely not have minorities and likely not be next to zoning districts that allow liquor stores. Does that seem like a good hypothesis? Any tips? 

[glawson]

I don't have a lot of experience with stats packages, but thought I'd mention there is a OSS counterpart to SPSS, which is PSPP (catchy, huh?).

http://www.gnu.org/software/pspp/

And there is NASA's World Wind, which has a lot of promise but which is a bear to get running on Linux (the "new" Java version). I have some more GIS bookmarks at home; I'll look through those tonight.

http://worldwind.arc.nasa.gov/

Greg
------------------------

[Seth Galitzer]

SPSS and ArcGIS are the industry standard when it comes to statistical
and geographic analysis. As far as I know, there is no OSS project that
comes close to touching either.

Seth

[Joseph Kearns]

About R, there is an unofficial gathering that the Geography Dept has
in Lindley Hall. I am still trying to find a way to fit it in my
schedule. The following is a post from their facebook page--

KU Geography useR! group :: A small group of us (Lisa Rausch, Lynette
Dornak, & Joel Plummer so far) are getting together to learn and
explore the use of the R statistical computing package. R is a free
and open source software package similar ...to Matlab. It is a very
powerful analysis tool that can be used to perform data manipulation,
simple and advanced stats, plotting/graphing, image analysis,
geographic analysis, and much more.

This is not a class or seminar, just a group of geo-geeks hoping to
learn more by working together. If you're interested in learning R, or
you're just curious, feel free to show up or drop any of us an email.
Right now we're planning to meet Mondays @ 3:30 in the lunchroom.

[Steve Nordquist]

Seth...on what features, say you? Do Federal agencies still use them
to produce their statistical abstracts? AutoDesk, Google Earth, and
10 other CAD firms surely have their contender GIS software, as well
as statistics software to back them up, not least Mathematica or Maple
with some polish on the GIS or data bounds. OSS would handle these
with disparate packages, having no particular incentive to send BD+RW
around...unless multisession works really well.

> How ya do?  I'm working on a Master of Urban Planning @ KU. My professors
> speak with certitude that SPSS and ArcGIS are the only options when it comes
> to statistical and geographic analysis. The Linux/Gnu alternative appears to
> be GRASS GIS and R statistics. Does anyone have experience using these
> software packages?

freshmeat.net + 'statistical' ~> mineral chalk

> As a side issue, any tips on crafting a hypothesis? For my purposes, I have
> to use data from the census or similar source. The question has to be
> proven, or not-(I'm confused about the null hypothesis) using census data. I
> had previously thought of what I consider to be a great question. Does the
> greater occurrence of liquor stores in minority neighborhoods lead to
> increased consumption of alcohol? Prima facie this phenomena is the result
> of zoning. More affluent neighborhoods will likely not have minorities and
> likely not be next to zoning districts that allow liquor stores. Does that
> seem like a good hypothesis? Any tips?

See Gladwell?
Well, finding neighborhoods with no zoning or transparent zoning (e.g.
the board checks the applicant's credit and calls the zoning and
permits on that...which I think I heard of in W. Colorado) to test the
hypothesis sounds problematic. Similarly (and as often called out in
The Economist) counting the number of indigent customers in an area
with census data (and the analysis published ~3yrs. later) has its
documented issues and active dissidents. It can also be contended
that it's not the white guys and Indonesian Christians (minority,
right?) getting drunk but their Lvl3 Fortress Guards in Dwarf Fortress
(hence Appendix 4, people getting effed up with intent but no visible
means.)

OMG. Minority Liquor Entrepreneur + I feel lucky ~>
http://books.google.com/books?id=DtjSpIRhNNMC&pg=PA217&lpg=PA217&dq=minority+liquor+entrepreneur&source=bl&ots=W5M2Kj54bq&sig=wZvj9XloLHYL1knn94Sx53zwTdI&hl=en&ei=7si8Sv3OG4u4MPuohM8K&sa=X&oi=book_result&ct=result&resnum=1#v=onepage&q=&f=false
...and the map at the bottom leaves out Coastal Africa.
...it says the Koreans sell wigs.

[Paul Johnson]

John Gary wrote:
> How ya do? I'm working on a Master of Urban Planning @ KU. My
> professors speak with certitude that SPSS and ArcGIS are the only
> options when it comes to statistical and geographic analysis. The
> Linux/Gnu alternative appears to be GRASS GIS and R statistics. Does
> anyone have experience using these software packages?

I user R lots. See

http://pj.freefaculty.org/R/Rtips.html

and

http://pj.freefaculty.org/stat/StuffWorthKnowing.pdf

R has many addon packages besides GRASS for work on spatial analysis.

GIS is nice for data management and making pictures, but the statistical
analysis part still has to be done with some real statistical software.
And I assure you SPSS is not it.

You might consider installing R and using the RSiteSearch() function.
It will scour lots of stuff.

I'd also suggest you study the list of packages here:
http://cran.wustl.edu/

(or at any CRAN mirror you can find by going to r-project.org).

>
>
> As a side issue, any tips on crafting a hypothesis? For my purposes, I
> have to use data from the census or similar source. The question has to
> be proven, or not-(I'm confused about the null hypothesis) using census
> data. I had previously thought of what I consider to be a great
> question. Does the greater occurrence of liquor stores in minority
> neighborhoods lead to increased consumption of alcohol? Prima facie this
> phenomena is the result of zoning. More affluent neighborhoods will
> likely not have minorities and likely not be next to zoning districts
> that allow liquor stores. Does that seem like a good hypothesis? Any
> tips?
>
> >
>

--
Paul E. Johnson email: paul...@ku.edu
Dept. of Political Science http://pj.freefaculty.org
1541 Lilac Lane, Rm 504
University of Kansas Office: (785) 864-9086
Lawrence, Kansas 66044-3177 FAX: (785) 864-5700

[jamesmi...@googlemail.com]

On Fri, Sep 25, 2009 at 7:41 AM, John Gary <joh...@ku.edu> wrote:
> be GRASS GIS and R statistics. Does anyone have experience using these
> software packages?

Howdy,
there are a bumch of groups on irc.freenode.net and online for grass.
I have installed grass and some others http://opensourcegis.org/
but have not gone into depth.

>Does the
> greater occurrence of liquor stores in minority neighborhoods lead to
> increased consumption of alcohol? Prima facie this phenomena is the result
> of zoning. More affluent neighborhoods will likely not have minorities and
> likely not be next to zoning districts that allow liquor stores. Does that
> seem like a good hypothesis? Any tips?

Do you have the location of these stores?
Can you get any data on consumption?
It think data collection is going to be difficult here.

mike

[glawson]

I just ran across this link to the SOFA Statistics application; I hadn't previously been aware of it.

http://www.sofastatistics.com/home.php

Greg
-----------------------------
.

----- Original Message -----
From: "John Gary" <joh...@ku.edu>
To: kul...@googlegroups.com
Sent: Friday, September 25, 2009 12:41:29 AM GMT -06:00 US/Canada Central
Subject: [KULUA] statistical packages, gis and hypotheses