2005 Statistical Highlights

Four Exciting Developments

By Frank LoPresti

SPSS Output Management System

SPSS Output Management System (OMS) provides an output control function very much like a series of "If" commands; they are turned on and off during the session. Using OMS, you would first create several output files, each with its own format: HTML; XML; SPSS Dataset; or delimited text. You can then choose what SPSS statistical commands will trigger the writing of output to the various files you created.

For example, you could open a file called "anova_out.sav" and, with OMS control, append the output to that file whenever an anova command is run. Since the file is formatted as an SPSS file, the statistics will be written as data, and the tables of Sums of Squares, F, etc., will all be saved as data. Say you have 20 years of weekly data and run similar anova commands on the 20 year groups; the file "anova_out.sav" would have 20 lines of data with the statistics from each of the anovas. The variables in this dataset would be statistics such as Sums of Squares or F—the various statistics created by the anovas.

Visit UCLA's OMS website, http://www.ats.ucla.edu/stat/spss/faq/oms.htm, for a variety of command syntax examples and other useful information. OMS is included with SPSS version 12 and above; NYU has a site license for SPSS version 13 (and previous versions, if required). See http://www.nyu.edu/its/statistics/sas_spss.html for information on acquiring SPSS from NYU Information Technology Services.

UCLA Academic Technology Services' Website

UCLA's Academic Technology Services website cited above (http://www.ats.ucla.edu/stat/) is a rich, public resource for SPSS, SAS, and Stata users, with helpful documentation and links.¹ It also includes a wonderfully useful table entitled "What statistical analysis should I use?" (see figure 1). Using the number and type (categorical or not) of dependent and independent variables you have in your model, you can find the recommended statistical test, and links to examples of how to do the test with SPSS, SAS, or Stata. A discussion of the results and references for further study are also available.

Figure 1. UCLA's Academic Technology Services' website offers many useful resources, including this statistical analysis table.

Other resource pages, many of them cross-linked to the indices, include code fragments for advanced users, statistical papers and bibliographies, and organized sections of links dealing in great detail with statistical models. For example, in the "What statistical analysis should I use?" table mentioned above, you could click on the SPSS link for the Repeated Measures test. You could then download the sample data, follow the instructions for running the GLM (General Linear Model) command, then compare your results to those shown on the page. There is also a discussion of the results and links to other code and documentation. The site provides information on the prerequisite knowledge you would need, or, for more advanced statisticians, confirmatory details on the parameters calculated.

MIT'S OpenCourseWare

OpenCourseWare (http://ocw.mit.edu/) is "a free and open educational resource for faculty, students, and self-learners around the world" that provides MIT lecture notes, assignments and tests for many courses. Courses useful to quantitative researchers range from introductory to advanced statistics. At the elementary end of the spectrum, for example, is Statistics for Applications, a "broad treatment of statistics." At the more advanced end is Introduction to Modeling and Simulation, an "overview of modeling and simulation tools, as well as case studies in modeling and simulation."

MIT President Susan Hockfield states that "through MIT OpenCourseWare (OCW), educators and students everywhere can benefit from the academic activities of our faculty and join a global learning community in which knowledge and ideas are shared openly and freely for the benefit of all."²

Figure 2. MIT's OpenCourseWare website, http://ocw.mit.edu/, offers public access to hundreds of online MIT courses in statistics.

OCW, which is funded jointly by the William and Flora Hewlett Foundation, the Andrew W. Mellon Foundation, and MIT, was launched on April 4, 2001. Its mission is to expand access to information to students and teachers around the world, and to counter the trend towards the privatization of knowledge. The site currently offers more than 500 courses, and plans to include 1800 courses by 2008.

The MIT site, along with the UCLA statistics site discussed above, are good examples of the level of maturity the Internet has reached. These sites are not merely academic papers published on a university's statistics website. They were not simply found by Googling a topic of interest and getting one treatment of a subject out of context. They are organized archives, rich with context, which allow you to step backwards or forwards in a question to learn important prerequisites or expert uses. Statistics scholars would be well served to visit these sites often.

Minitab Release 14

Minitab, a well-respected 30-year-old statistical package, has just released a new version. It was originally developed at Penn State with contributions from many talented statisticians. Minitab has a feel similar to SPSS or Stata and is used at NYU for several introductory statistics courses. To take a tour of the new release, which offers increased graphics and graphing capabilities and a customizable interface, visit the Minitab website at http://www.minitab.com/products/minitab/14/ (see figure 3).

Figure 3. Tour Minitab 14 at http://www.minitab.com/products/minitab/14/.

One immediate advantage of Minitab over SPSS, Stata, and SAS is the cost. A semester rental of the student version is available on the Internet for only $30 and is sometimes included with textbooks.³ By comparison, NYU students can purchase:

• Stata for $130 (http://www.nyu.edu/gsas/dept/politics/datalab/)
• SPSS student version for $200 (http://www.bookstores.nyu.edu/computer.store/)
• SAS for $75 per year after an initial fee of $175 the first year (http://www.nyu.edu/its/statistics/sas_spss.html)

All of these packages are also available for use at the ITS Third Avenue North computer lab (http://www.nyu.edu/its/labs/third/). NYU presently has a site license that allows Minitab to be installed at no cost on University-owned computers. Minitab does not allow for the easy transfer of datasets between statistical packages, so additional software that facilitates this process (such as StatTransfer) should be installed on the computer to complement Minitab. Minitab does have certain limitations that prevent it from being as useful as SPSS, Stata, and SAS for packaging, managing and documenting data.⁴ Nonetheless, the fact that it is easy to learn Minitab and transfer that knowledge to other programs makes it a good choice for introductory courses.

If you have questions about any of the resources described here, please send e-mail to: frank.lopresti@nyu.edu.

Author Biography

Footnotes