Content area
Full text
Editor's corner
This issue has a review of the sister packages SigmaPlot 2000 and SigmaStat2, now owned by SPSS. These packages are aimed at students and claimed to be of intermediate level, i.e. less complex than SPSS or SYSTAT. SigmaPlot is distinguished by its powerful graphics; and SigmaStat by the advice and explanation of the underlying statistics given with all procedures. Due to inefficiencies in my time management, I am the reviewer as well as the software editor for this issue. The review machine is a Mac G4 with Virtual Memory software installed with 1.5Gig assigned disc space.
SigmaPlot
The major advantages of SigmaPlot lie in the number of different kinds of plot that can be produced and in the amount of tailoring that is possible for every aspect of the graph. In reviewing the package I had several standards in mind that I use frequently to produce graphics, specifically: EXCEL, STATVIEW and JMP-IN. The latter two are originally Mac products and so, obviously, have far superior interfaces even when ported to the clumsy Windows environment.
Functionality
It is on functionality that SigmaPlot scores well, so this will be the start point. The possible graphs include the usual scatter plots and regression plots, 2- and 3-D bar charts and box plots (but without notches, etc.), as well as survival analysis and more complex 3-D plots including contours and waterfalls. All except the waterfall are also available in EXCEL.
The regression plots were impressive in that one can have either multiple X-Y pairs, or many Ys with same X, or many Xs with same Y. One may then fit any of the usual functions to each curve: linear, polynomial, log, exponential logit, etc. SigmaPlot then provides the best fitting curve, with associated confidence limit curves. Where there are several functions on one plot, one may obtain distinct functions for each curve (e.g. linear for one and log for another). JMP-IN can do most of this but does not include logistic or normal z-scores. STATVIEW can also do most of this but again the number of transformations is more limited.
Like the other packages, SigmaPlot will calculate confidence levels from the data and has two very useful additional features. Firstly, one can specify upper and lower...