Statistical Software at UVa

1
Statistical Software at UVa

• By Tim F. Jost Tolson
• ITC Research
• Computing Support
• Res-Consult_at_Virginia.EDU
• Phone [phone] F Fax[phone]
• http//www.itc.Virginia.edu/researchers

2
Topics

• Background
• General Advice
• Statistical Software Packages at UVa.
• Which one should I use?
• Getting Help
• Where to from here?

3
General Advice

• The mark of a trained mind is the ability to
  expect no more precision from a subject than is
  inherent to its nature. -Aristotle
• The government is very keen on amassing
  statistics and preparing wonderful diagrams, but
  you must never forget that everyone of these
  figures comes in the first instance from the
  village watchman, who just puts down what he damn
  well pleases. Quoted in Sir Josiah Stamp, Some
  Economic Factors in Modern Life_ From the SHAZAM
  User's Reference Manual.
• Numerical representation statistical packages
  store categorical numbers (integers) as real
  numbers, despite what they may display.

4

• If all you have is a hammer, everything looks
  like a nail! ?Excel is not the right tool for
  statistical analysis.
• Garbage in Garbage out. Check and re-check
  your data using frequencies, descriptive
  statistics and inter-ocular test.
• Use labels for variable names and categorical
  data values.
• Annotate and document along the way.
• Getting a copy http//www.web.Virginia.edu/rescom
  p/ldb/swdb.asp

5
SPSS

• Used to be Statistical Package for Social
  Sciences. One of the first statistical packages.
  Originally written in Fortran.
• Low learning curve easy to use (and misuse)
• Powerful. Can do both menu and command programs.
  Has built-in macro language.
• Modules available vary by platform Windows,
  Macintosh and Unix (IBM-RS/6000 and Sun).
• Most widely used package at UVa and many
  educational and social science centers.

6
SPSS Modules

• Windows, version 10.1 Base, Regression, Advanced
  Models, Tables, Trends, Exact Tests, Missing
  Values
• Macintosh,Version 10.0 Base, Regression,
  Advanced Models, Tables, Trends.
• Unix (IBM Sun) Version 6.1.4 Base,
  Regression, Advanced Modules, Categories, Tables.

7
SPSS Key Features

• Well integrated into Windows and Macintosh
  platforms.
• Spreadsheet view of data
• Graphing capabilities easy to use and good.
• Can start with point and click with mouse and
  easily move to building command strings that can
  be saved to files for repeated use later.
• Easy and good cross-platform and
  cross-application file sharing.

8

• Easy to document and annotate data file.
  Variable and value labels and dictionary.
• www.itc.Virginia.edu/research/spsshelp.html

9
SAS

• Used to be Statistical Analysis System, then
  Strategic Analysis System, now just name. Early
  alternative to SPSS. Originally written in
  Fortran on IBM Mainframe. Version 6 was a
  complete re-write in C!
• Steeper learning curve than many other packages,
  but also more powerful and flexible. Still
  relatively easy to misuse.
• Powerful. More programming, command language
  oriented. Has built-in macro language. Many
  modules, parts, add-ons. Very strong data
  management.

10

• Modules available vary by platform Windows,
  Macintosh and Unix (IBM-RS/6000 and Sun).
  Macintosh version frozen at 6.12
• Many books and manuals available.
• Widely used package at UVa. Used by
  administrators for paychecks, reports, etc.
  Widely used in industry-MCI, banks, Census Bureau
  and across a great variety of disciplines from
  Anthropology to Systems Engineering.
• Huge Users groups SUGI and VASUG

11
SAS Modules

• Way too many modules to list here. See
  http//www.itc.Virginia.edu/research/sashelp.html
  modules
• Windows, version 8.2
• Macintosh,Version 6.12
• Unix (IBM Sun), Version 8.2.
• Will be adding Linux and Windows NT server this
  fall.

12
SAS Key Features

• Strong, flexible data step for data management
  and manipulation.
• Not the default, but can get a spreadsheet view
  of data and do data entry therein.
• Graphing capabilities great, but harder to use.
• Requires learning syntax and programming/syntax
  rules..
• Has built-in modules for financial datasets, such
  as CRSP, Compustat, etc.

13

• Variable labels easy to use, value labels harder.
  Can make SAS dataset unreadable if not properly
  done. Separate files for data and user-applied
  data formats (labels).
• Has many good manuals written by users to get
  started.
• www.itc.Virginia.edu/research/sashelp.html

14
S-Plus

• Based on S Unix statistical program developed
  by Bell Labs. Insightful Corporation
• Kind of a cross between SAS and SPSS
• Pull-down menus combined with powerful command
  language.
• Ability to write your own functions and integrate
  easily into S-Plus.
• Data management, sharing is improved with Version
  6

15
S-Plus Key Features

• Excellent, flexible graphing capabilities.
• Updates and integrates latest statistical
  procedures frequently.
• Does requires learning syntax and
  programming/syntax rules.
• Has more of a matrix/vector approach to data
  files than traditional case/variable approach
• www.itc.Virginia.edu/research/splushelp.html

16

• Windows version called 2000, Version 6 out.
• Unix version 5.1, version 6.0 available soon. On
  IBM, SGI, Sun, and Linux
• No Macintosh version.
• Additional speciality modules not licensed by
  ITC, including tie-in to ESRI/ArcView and a
  wavelet analysis toolbox.
• http//www.insightful.com/products/modules.html

17
Stata

• Powerful graphical, statistical, and data
  management environment.
• Easy to use pull down menus.
• Powerful. Wide variety of statistical techniques
  from t-test to time series, means to matrix
  manipulation all in one integrated package.
• Has two size versions Small Stata and
  Intercooled

18
Stata Key Features

• Fast, easy to use.
• Programmable, extendable write your own
  functions and add them in easily. Stata staff
  and user add new features frequently
• Windows version 7 (we have version 6 now, getting
  7 for public labs this spring.
• Macintosh Version 7 and version for OS-X

19

• Unix version available from Stata, not licensed
  here at UVa. IBM, Sun, and Linux.
• GradPlan for getting copies.
• http//www.itc.Virginia.edu/research/statahelp.htm
  l

20
Minitab

• Minitab, Inc. at State College, PA. Like SPSS
  and SAS origins are link to a college or
  university.
• Integrated statistical and graphical analysis
  package.
• Widely used for teaching statistics and analysis
• Specially focused on quality improvement analytic
  techniques.

21
Minitab Key Features

• Quick and easy to learn.
• Good graphing capabilities.
• New ActivStats for learning and understanding
  statistical procedures.
• http//www.itc.Virginia.edu/research/minitabhelp.h
  tml

22
Lisrel/Prelis

• Program for confirmatory factor analysis and
  structural equation modeling.
• Advanced statistical procedure requiring advanced
  statistical background and training.
• Available on Unix (node1, node2, node3.unix)
  and Windows.
• http//www.itc.Virginia.EDU/research/lisrelhelp.ht
  ml

23
Amos

• Easy to use structural equation modeling program.
• Stands for Analysis of MOment Structures and is a
  product of Smallwaters, Inc.
• Distributed by SPSS and can read SPSS system
  files.
• Has graphical user interface as well as command
  syntax. Some options only through commands.
• Windows only version. No Unix or Macintosh.
• http//www.itc.Virginia.EDU/research/amoshelp.html
  

24
BMDP

• BioMedical Data Processing program
• No longer available, no updates. Bought by SPSS,
  Inc. years ago.
• On blue.unix only, will be gone when it stops
  running because of operating system changes.
• Individual modules that have to be invoked
  separately. Cannot mix data steps and different
  analytic procedures.

25
ESRI/ArcView

• Desktop Geographic Information System (GIS) suite
  of programs.
• Distributed by us here at RCSC
• Primary support by GeoStat Center in Alderman
  Library.
• Many modules, options, add-ons, etc.
• http//www.itc.Virginia.edu/research/arcview.html

26
Other Packages

• Systat Statistical and graphical analysis
  package similar to Stata or S-Plus, made by SPSS,
  Inc. Analytic procedures of BMDP have been added
  to it. Not licensed by or available from ITC.
• http//www.spss.com
• SigmaPlot Graphics package, distributed by SPSS,
  Inc. Not licensed by or available from ITC.
• http//www.spss.com
• SudaanSUrvey DAta ANalysis for multi-stage
  sample designs.
• http//www.rti.org/sudaan/

27

• R is GNU S', a freely available language and
  environment for statistical computing and
  graphics which provides a wide variety of
  statistical and graphical techniques linear and
  nonlinear modeling, statistical tests, time
  series analysis, classification, clustering, etc.
• http//lib.stat.cmu.edu/R/CRAN/
• http//www.R-project.org

28
Which one should I use?

• What are your colleagues here and in your field
  using?
• What statistical procedures do you need? What is
  cutting edge in your field?
• What platform Windows, Macintosh, Unix?

29

• Balance among
• ease of learning and use
• Power, expandability, flexibility
• Data management and sharing data with other
  statistical packages.
• Innovativeness
• Graphical capabilities
• Cost!

30
Getting Help

• Statistical Computing Consultant at the Research
  Computing Support Center
• Help you help yourself from start to finish.
  From I have these data (or I want to collect
  these data) to complicated statistical analysis
  or data manipulation.
• We can help move data from one program/filetype
  to another and between computer platforms.
• Will try to help with packages we dont license
  or support.
• GeoStat Center in Alderman Library
• http//fisher.lib.Virginia.edu
• Phone [phone]

31

• Statistics Department does for-fee consulting and
  programming.
• http//www.stat.Virginia.edu/consulting.html
• Phone [phone]
• Health Evaluation Sciences
• http//hesweb1.med.Virginia.edu/
• Center for Survey Research
• http//minerva.acc.Virginia.edu/surveys/

32

• Where to from here?
• Visit our website. Use one of our tutorials or
  built-in tutorials.
• Schedule an appointment with statistical
  computing consultant.
• Visit Librarys GeoSpatial and Statistical Center
  (GeoStat) in Alderman Library.
• Take our SAS or SPSS Workshop dates in October
  or self-paced on the web .
• Thank you for coming. Please return for our next
  talk, Wednesday, October 17 on Web Databases and
  Tools at Noon here.

33
Coming Soon

• Labview training next week
• SPSS and SAS Computer Workshops in October.
  Introductory. 2 nights for 2 hours from 600 to
  800 PM
• Remaining Brownbag Talks here at RCSC
• Web Databases and Tools, Wed., October 17 at Noon
• ESRI/ArcGIS by Mike Furlough, director of Geostat
  Center here on Wednesday, November 14 at Noon.

34
Conclusion

• Do not put your faith in what statistics say
  until you have carefully considered what they do
  not say. -William W. Watt
• Then there is the man who drowned crossing a
  stream with an average depth of six inches.
  -W.I.E. Gates