There is but one argument in computer- assisted reporting that rivals that of PC versus Mac. It gets blood boiling and inspires people to write opinionated tomes on the NICAR-L listserv. It's the debate over the merits of SAS versus SPSS. Users of the statistical programs each trumpet the advantages of their pet program ("SPSS is easier!" "SAS is more powerful!") but occasionally one can hear them muttering over the disadvantages inherent in each: "SPSS can't do what I want it to." "SAS is just too expensive." So which is the better program?
The experts say it really depends on your needs in your newsroom. The pluses According to most, SPSS is usually best for beginners in CAR. One important factor is its relatively low cost - about $1000 for commercial use, half that for academic use and even less for individual students (both graduate and undergraduate versions are available). It also requires less complex commands and is quick to learn.
Philip Meyer, Knight Chair and professor of journalism at University of North Carolina-Chapel Hill, is confident in the program's usefulness for CAR work: "SPSS will handle every statistical problem that I can think of that a journalist might encounter. Including some weird-science stuff like logistic regression and factor analysis." Finally, SPSS is easy to understand Meyer calls this factor "transparency." "Transparency means being able to see right through the complicated computer operations as though they weren't even there.
All you see is your input and the output. Another expression for transparency is 'ease of use, he says. "Every computer system designer has to choose between transparency and flexibility. SPSS leans toward transparency, SAS toward flexibility." Which brings us to SAS's main advantage: it can do more than SPSS, a benefit that the SAS users tout, though their voices may be drowned out by those of the more numerous SPSS users.
SAS can bring in a number of different data types, and it creates detailed logs each time programs are run, making it easier to dig out any bad data. Because of its ability to handle very large data sets, oftentimes SAS is used by the professional agencies that supply reporters with data. Steve Doig, professor at the Cronkite School of Journalism at Arizona State University, describes SAS as "extremely powerful and full-featured.
"You just won't outgrow this, Doig says. The minuses Despite their preferences for one or the other, users of both programs are honest about the disadvantages inherent in each, which are sometimes related to the advantages. SAS's power, for example, comes at a cost - financial as well as the time required to learn the more complex procedures. "You have to write programs," Doig says. "There are pulldown menus, but they don't do many of the programming functions that the SPSS interface does." Also, the monetary cost can be prohibitive, especially for smaller newspapers or television stations.
A license must be purchased to run SAS, rather than just a neat, cellophane-wrapped package of software. A first-year license fee for a basic SAS program will run you $2,200-2,300. The yearly renewal cost, though, is much cheaper - about half the initial amount. Any stats modules licensed on top of that basic program cost an extra $1,100-1,200. But the base program, which Doig says can handle chi-squares, frequency tables, etc., should be sufficient for most work.
Expense and complexity both limit the number of CAR reporters who become SAS users, which might be a problem if you want to ask questions about your work. But Doig says there is help available: "We few SAS zealots are so tickled when someone does want to use it that we're generally happy to chime in." SPSS is cheaper, but has its own drawbacks, It can't handle large data loads as SAS can, nor can it do complex file-management tasks as well.
A few NICAR-L complaints about SPSS have included its limited ability to handle text fields, a lack of some key diagnostics for basic statistical procedures and less efficient data cleaning than SAS. Though he recommends SPSS for beginners in CAR, Meyer acknowledges its shortcomings in terms of flexibility. "Bottom line: you can do especially file-management tasks, with SAS," he says. "For those with the guts to stick out the SAS learning curve, it will last them for- ever.
Those fainter of heart should start with SPSS. They can always get an outside contractor to do the slice-and-dice chores." What's right for you? In the end, it all boils down to a few basic questions: a) How much intensive statistical work do you need to do? b) How much statistical experience do you have? How much can you or your organization afford? If you answer "A lot" to all three, SAS may be your best bet.
If it's "Not much" across the board, SPSS is probably more suitable. Any combination of the two responses means you need to weigh the pros and cons of each and maybe post a question or two to NICAR-L in hopes that a fervent SAS or SPSS user can sway you to one side or the other. Jessica Larson can be reached by e-mail at jessica@nicar.org