TechWeb

To Build A Survey

Sep 24, 2005 (01:09 PM EDT)

Read the Original Article at http://www.informationweek.com/news/showArticle.jhtml?articleID=171200421


In the academic research field, speed and efficiency are almost as important as they are in the business world. At the Research Institute on Addictions (RIA), which is part of the University of Buffalo, between 30 and 40 long-range behavioral studies on gambling, drug use and alcohol are being conducted at any one time. To operate those projects as seamlessly as possible -- not an easy task given the complexity of the investigative assignments -- the institute needed some help.

According to George Gogos, head of computer programming at the institute, the implementation of software from SPSS has helped RIA move its studies more efficiently from drawing board through completion.

"Back in the day, after a project got its grant funding, it often took almost a year until we were ready to go into the field," Gogos says. "Now, we can get up and running within two or three months."

The RIA's studies are composed of enormously elaborate questionnaires delivered via various methods. Subjects might sit down with pencil and paper as they would with a standardized test. They might click through surveys published on the Web (a function the institute only acquired with the SPSS implementation). Interviewers might come to the subjects' houses with laptops and conduct surveys face-to-face. Or interviewers might sit at workstations in the RIA office and conduct surveys over the telephone. Most of the institute's clients are psychologists or bio-medical researchers, and their studies are directed at large samples of people -- often enough numbering in the thousands. One ongoing study, which has looked at the behaviors of married couples, began the moment the brides and grooms received their marriage licenses. After nine years, the study is currently surveying those couples' children.

The complexity does not end there -- further wrinkles add to the difficulty of creating the studies and getting them out the door and into the field. Researches must identify a sample -- a huge list of potential subjects, randomly generated. The sample requires a significant amount of management time and energy, as people on the sample list are convinced (or not) to participate in studies. The surveys themselves contain questions that must be phrased precisely so as not to affect responses, and the questions must come in a specific order, for similar reasons. Often enough, if a subject responds to a query with a certain answer, the response will trigger another subset of questions -- called a "loop --and, many times, loops will occur within loops. Because of this complexity, the questionnaires must undergo a series of tests before they go out into the world, and this is one of the areas in which SPSS has helped, according to Gogos.




"The script SPSS is using allows us to rapidly create the questionnaire and test it almost in one sitting," he says. "That whole testing and debugging cycle -- we can get things fixed earlier." Called Quancept, the SPSS scripting technology is a proprietary language that has made Gogos' job a bit easier. In the past, the IT department had to become much more involved in that testing process, going into the system and helping RIA researchers work out the glitches. Sometimes those glitches -- mostly having to do with skip logic and all the loops that occur inside the questionnaires -- wouldn't be corrected until after the surveys were in actual use. "Looping errors are usually tough to debug," Gogos says, but, he added, the SPSS product is better than the RIA's previous technology at letting programmers where in loop the error has occurred.

The institute first installed SPSS in April 2004 and deployed it in its first study, a relatively simple one, five months later. Since then, the institute has slowly phased out an older DOS-based interview and survey-generating product, which suffered from a few drawbacks. Gogos won't reveal the name the vendor. "It was a great product for its time," he says, but it had aged substantially over the years.

Because it was DOS-based, the scripting technology was cumbersome. Also, it was not automated. After researchers completed a survey, the IT department had to manually extract the data from each individual workstation and laptop. With SPSS, the data is removed each night automatically and placed into one cohesive file in a data warehouse, where it is "scrubbed and checked," Gogos says. RIA researchers can then easily view the data via the Internet -- another new capability offered by SPSS -- and obtain a "thumbnail sketch" of a study's progress. Furthermore, Gogos and his team have been freed up from those workflow demands, and they can concentrate their time on other work, such as helping to create questionnaires and prepare studies.

Gogos wouldn't divulge how much the institute paid for its 25 SPSS licenses. Nor would Rich Rodts, director of higher education sales at SPSS, who said that pricing varies widely from customer to customer, depending on the package of bells and whistles that each user might need. Higher-ed business at SPSS is significant. Rodts said about 90 percent of American universities use SPSS in one way or another, many of them for research projects similar to those carried out by RIA.

So far, at the University of Buffalo, RIA is using SPSS in only about ten of its ongoing projects. In an additional five studies, the institute is employing SPSS "in the background," by which Gogos means that researchers are using the software only to define a study's data requirements, rather than actually deploying the questionnaires via the SPSS interface.

Also, the institute has not yet implemented call-management software from SPSS. For telephone surveys, the RIA still stores its sample lists on the old DOS-based product. A homegrown bridge program -- written by Gogos in PERL using a connectivity protocol designed by Microsoft -- then transfers the information from the call management software to SPSS, which tracks and runs the surveys from that moment on. "We logged over 20,000 calls in just a few weeks, and the bridge has worked flawlessly," Gogos says.