STATISTICS

Statistica 8

12 June 2007

StatSoft

Reviewed by Felix Grant

Like many product lines at the top of their tree, these days, Statistica now comes in different weights and configurations. This review is of the early release 8 Desktop (or single machine) Statistica Base version, with immediately available advanced modules; the Enterprise variant is still to come, ‘in a couple of months’ at the time of writing, as are updated data mining options.

There are no radical shocks in this upgrade: development is substantial but evolutionary, strengthening and extending within the existing skin rather than reshaping. There has been a degree of reorganisation, but discretely so; most users won’t notice it, and certainly shouldn’t find workflow disrupted. New features are, for the most part, placed into existing blank spaces within existing structure and layout. A check box for botched runs appears in previously blank grey space next to the variables button of two variable DoE designs, for instance, and a centre points variable then adds itself to the selector when the button is clicked. A counter example is expansion and enhancement of the ‘By Group’ interface, which has shifted to tab in the specification of analyses, though the older (script-based) method still works. Some areas have to do with size or scope of handling, applicable to extreme or unusual situations - nonparametrics up to a million cases, for instance, or provision for 500 sort keys in the spreadsheet

There is a good spread of onboard additions to the statistics capability, from refinements for basic descriptors and tests to expansion of quality control tools in several dimensions, and additional modules see extension, expansion or repackaging. Statistica Automated Neural Networks package offers not only a boost in capability but a considerable step forward in usability too - to which I plan to return in more detail, along with the other advanced modules, when all are available.

Front-end and preparatory support for the statistics functions have kept pace. The spreadsheet acquires a hefty expansion of its formulae repertoire to calculation of compound or derivative variable values within cases, and they can call on cell grid references. A number of flexible methods are provided for intelligently dealing with decisions about the handling of data within the sheet such as excluding (or otherwise dealing with) cases which exceed a user specified level of incompleteness, duplicate data, show variation below specified levels, and so on. Importing data coming from text files is both quicker and more intuitive, and merging files is a smoother business. Outliers can be handled in various ways and optionally recoded.

Usability and integration with the larger working and communication environments have for a long time been a strength of Statistica, and continue to develop in this release. Integration with Microsoft Office includes opening Excel workbooks directly as a replacement for Statistica’s own worksheet; outputting reports to Word format documents will be popular, though the ability to drag and drop report content (text, tabular or graphic) to a chosen point within any appropriate output handler (a different word processor, a desktop publisher, a spreadsheet, a graphics package) is a richer and more flexible alternative. For output to bitmap file pending later use, graphics (which benefit from several innovations not mentioned here) can now be saved at a user specified resolution. Locking has been introduced to avoid inadvertent changes to graphics or worksheets in a multiuser environment.

There’s much more, but at the bottom line this is an upgrade which will pay for itself in very short order through both enhanced capability and more deeper, more intuitive usability.

Click here to find out more