SPSS/PC: A Quantitative Historian’s Dream or Nightmare?

Anyone frustrated by the delays and breakdowns of mainframe computers, has experienced a recurring, seductive vision: If only the same powerful pro­gram package would run on one’s own friendly microcomputer! In the last sev­eral months, this dream, born of anger over down-time and interminable turn­ around, has come one step closer to reality: With much fanfare, SPSS has introduced a microversion of its main­ frames “Statistical Package for the Social Sciences” for the IBM PC. In the words a promotional letter: “The overwhelming demand from our users for a statistical package on an IBM PC has led us to the development of SPSS/PC.”¹ Historians ought to have been curious about this announcement, since SPSS is the package most used by them on the mainframe. Moreover other statistical software for microcomputers has to this point fallen considerably short of their needs.²

Shoehorning a large and complex sta­tistical package into a microcomputer was no mean feat. SPSS/PC therefore does have system and space require­ments, which will put it beyond the reach of most popular micros, used for word processing (such as the Apple IIs or the Kaypro): It needs an IBM PC (or XT), with at least 320K of memory, a hard disk, also a math-coprocessor (an 8087 chip) to speed up its calculations, and MS DOS. The hard disk is neces­sary, since the program comes on nine diskettes, each containing modules and subroutines, which need to be available concurrently. These are substantial hardware requirements, putting the necessary system into the top range of personal computers. But with the con­tinued memory/size expansion of mi­cros and the dropping prices, such ma­chine configurations are certainly with­in reach of departments and increasingly also of individual scholars.

In actual operation, the microversion of SPSS is remarkably similar to its big­ger sister. It can handle up to 200 varia­bles at the same time and disk space is the only limiter of the number of cases. In practice it runs relatively quickly for data sets of let’s say twenty-five variables and 1,000 cases, even if it has to switch modules along the way. The major syn­tax innovation is a period behind each command, since there is no longer a computer card ending which would automatically indicate the command end. Should one forget to include the period, the package prompts the user to supply it or other additional informa­tion.³ SPSS/PC must be run from the hard disk and cleverly requires a key diskette (as copy protection) to get it started each time.

Fortunately, the program comes with a tutorial, which allows the novice to try out its main features before creating his/her own data set. In general the operation is, for such a large package on a small machine, efficient and user friendly. Especially for those scholars, already somewhat familiar with one of the mainframe versions, the transition should be fairly easy.

The initial working impression of SPSS/PC is therefore positive, not to say exhilarating. Finally, the historian can enter his own data, check them through a FREQUENCIES command and run simple procedures such as CROSSTABS (with the same statistics as on the main­ frame) all in the same session at his desk in his own study! No need to punch cards any longer in a noisy and frantic users’ room at the computer center or to fight for a terminal. This favorable experience is reinforced by the interac­tive nature of SPSS/PC, which is a decid­ed step beyond the mainframe version: Not only are commands immediately displayed on the screen, but error mes­sages also flash up in an instant so that mistakes can be corrected without frus­trating waiting for a disappointingly thin printout of another “busted” run.

Moreover, the package has a help prompt that explains the basic features of any procedure so that the researcher can make required corrections without consulting printed instructions. The manual retains its accessibility for his­torians through a large “statistics guide” section that explains the purpose and limitations of different procedures on the basis of actual working examples.·

Furthermore, SPSS has kept many of its data transformation capabilities (so important to historians). It also offers a broad array of descriptive and analytical statistics from simple data listing and FREQUENCIES all the way up to FAC­TOR and CLUSTER analysis. Finally, it has added a PLOT procedure (which will even run without a graphics card) and has entered the exciting world of “log-linear analysis,” a promising statis­tical development for historians con­cerned with nominal and categorical data.⁴ The resulting table display is, given the limitation of screen size, at­tractive and easily printed in the format desired. On the whole the microversion of SPSS is therefore closer to the new SPSS-X release for the mainframe than to the earlier version, and in its interac­tive behavior, even a step beyond.

Skeptics are likely to ask: This seems too good to be true, what is the catch? The first several weeks of working with SPSS/PC did, indeed, reveal a number of shortcomings. To begin with, some power had to be sacrificed in the “down­sizing.” Therefore the DISCRIMI­NANT, MANOVA and SURVIVAL pro­cedures were dropped from the micro­version.⁵ Second, data files cannot be created within SPSS, but must be pro­duced by a text editor, such as the clumsy EDLIN (in DOS) or the fancier word processing packages like Microsoft Word. One needs to remember, howev­er, the SPSS will accept only standard ASCII files, a limitation that produces gibberish when not obeyed! Third, SPSS/PC will process merely one file at a time, which is a serious drawback for historians, since it means that they need to do their file merging, etc. in some other software environment or on the mainframe computer.

However, SPSS now also utilizes the matrixes created by simple procedures like crosstabulations as data for higher statistics such as factor analysis, a considerable convenience. Finally, the asking price for the copy-protected pack­ age is $795, a rather steep cost for an individual scholar. Initially, the pro­gram was not sold by buying clubs, although one would expect that with proliferation some discounts will be­come available (for instance if a store orders more than 10 copies, the price drops to the mid-500s). While some of these drawbacks (such as leaving out some procedures) are understandable, users ought to be warned more clearly about the need for a text editor.  And one would hope that the single file limi­tation and the communications prob­lems with the mainframe will be reme­died in future releases.

Is SPSS/PC therefore, the long-await­ed dream or yet another software night­mare? The answer to this question de­pends somewhat upon expectations and alternatives. While a microversion of BMPD has apparently come out, this biomedical package is rarely used by historians. The more frequently em­ployed SAS has the problem of being written in PLI, which is not available on IBM PCs and is therefore going to take a considerable time in coming up with a micro-conversion. Compared to the ear­lier packages (ABSTAT), SPSS/PC wins hands down.⁶ Its strength is with small­ to medium-size data sets of one- to two­-dozen variables. Those files SPSS/PC can handle quickly, efficiently and inter­actively. Really large data sets or huge numbers of variables are still better processed on the mainframe. To facili­tate such communication, SPSS does, fortunately, also contain an up- and downloading routine, called KERMIT.

Since historians are more likely to work with moderate numbers of cases and variables, SPSS/PC marks a decided advance in their working tools. While a final evaluation will have to await the experience of hundreds of hours of running time of all the intricate opera­tion, data-definition and procedure commands, the initial impression of SPSS/PC is quite favorable. The micro­version of this statistical package has the great advantage of disengaging quanti­tative methods from the big technology environment and of putting great statistical powers into the hands of the indi­vidual researcher. In many ways SPSS/ PC therefore promises to be more com­patible with the artisanal work style of historical craftmanship than its main­ frame counterpart.⁷

Notes

1. “SPSS/PC and SPSS/Pro Open SPSS to Micro World,” Keywords, 33 (1983), 2ff. It was actually first available on the DEC Professional 500. [↩]
2. D. K. Rowney, “The Historian and the Microcomputer,” Byte, 7 (1982), 166- 176; R. Jensen, “The Microcomputer Revolution for Historians,” Journal of American History, 14 (1983), 91-111; and M. H. Finefrock, “The Hard Software Decision” (part II of “Computers and the Historian”), Perspectives, 21 (1983). [↩]
3. M. J. Norusis, SPSS/PC for the IBM PC/XT (Chicago, 1984). [↩]
4. J. M. Kousser and others, “Log-Linear Analysis of Contingency Tables: An In­troduction for Historians,” Social Sci­ences Working Paper, 417 (1982) of the California Institute for Technology. [↩]
5. N. Nie and others, SPSS: Statistical Package for the Social Sciences (Chicago, 1975), 2nd ed. and SPSS-X Users’ Guide (Chicago, 1983). [↩]
6. R. McCaa, “Microcomputer Software Designs for Historians,” Historical Meth­ods, 17 (1984), 68-74 fn. 7 lists the relevant literature. [↩]
7. K. H. Jarausch, G. Arminger and M. Thaller, Quantitative Methoden in der Ges­chichtswissenschaft: Eine Einführung in die Forschung, Datenverarbeitung und Statistik (Darmstadt, 1984). [↩]