Systat 12
Systat is one of the long established names in data analytic circles, originating in the pioneering 1970s work of Leland Wilkinson at the University of Illinois, Chicago and uncompromisingly ploughing a distinctive furrow ever since. It is also one of the serious language based scientific tools in the market – a quick (admittedly not rigorous) check of peer reviewed articles in my bibliographic database for the past 12 months shows more citing Systat than any other equivalent product.
The statistical tool set is extended considerably, though solidly not flashily, in this release. There are new additions, expansion of existing provision, and some useful modifications well executed without significant disturbance to existing practice. There is some particular attention to areas of interest for industrial quality work and designed experiments (response surface methods being a prominent and valuable example), a common theme in current and recent analytic software development, but these aspects do not dominate.
The number of distributions available to fitting, plotting, the probability calculator, and various other constructs, has been expanded by almost a dozen new arrivals ranging from noncentral versions of t, χ2and F through generalised lambda to the intriguing inclusion of Benford’s Law. Generalised linear models see substantial fleshing out, a mixed models suite has arrived, series analyses gain nonparametric trend estimation and detection, an Anderson-Darling test makes its appearance in a spread of useful places, survival analysis is enhanced, and so on.
Equally important to many, especially new or prospective boarders, will be the continued attention to usability aspects.
A particular strength of Systat’s inheritance, and one which derives directly from Professor Wilkinson (an active researcher who has resumed a more prominently active relationship to the software), has always been the emphasis on a psychoperceptual approach to graphical data display exploiting multiple visual modalities. For a long time, however, that illuminatory capacity was hidden under a bushel of learning curves. My reviews of Systat described it as a powerful but 'fierce' product, a place where you were expected to know what you were doing and not be a crybaby, but since its acquisition by Cranes Software this has been softening. The competence is not being compromised, but the ease with which new arrivals can access it is progressively opening up. In some cases I have found that commands previously available only from the language command line are now available from mouse and menus; in others, already available techniques have been made easier to find and apply through new or rearranged interface option links. Users with an interest in exploratory practice will particularly welcome this continued evolution.
In more general ways, too, the same interface continues to move towards collective experience and familiarity with de facto conventions. The tabbed notebook structure, customisable to tiled or cascaded display; the worksheet and editors have become more like what you expect from elsewhere (and some editors have been retired, handing over their tasks to your own preferred alternative). There are still some idiosyncratic aspects, but they are now definitely the surprise exception rather than the rule.
Evolution, used above, is the operative word. There are no startling or unsettling shifts to destabilise those moving up from release 11 for the stretch of behind the scenes development, but migration from other environments is becoming easier all the time.
