Package Reports 54 Prism The Prism System is a computer package for statistical data analysis. It is being sponsored by the Working Party on Statistical Computing of the Royal Statistical Society, and is a development from GLIM. The latter was released in 1978 is now available at nearly 500 sites world-wide, including most British Polytechnics and Universities. Prism is an interactive program designed for the simple but flexible manipulation and statistical analysis of data. It is directive-based, each directive performing a specific task. Structures (which include scalars, vectors, strings, macros, labellings and n-way arrays) may be created, named, manipulated, displayed and deleted; actions are specified via pathwords, which can be combined in Englishlanguage-like phrases; space is allocated a n d recovered dynamically; multiple input/output channels (including transcript) are permitted. Full diagnosis of user faults, plus suggested remedies, are provided. Prism is arranged in 4 modules: [Kernel] [OL, r I ovl] IOR -I Each module contains approx. 10K lines of Fortran code. Any combination of modules that includes the Kernel can be used to form an executable program, so that, for example, the Kernel + GLIM-4 can serve as the replacement for GLIM-3, but there is also full compatibility between modules in the same program. The Kernel contains the housekeeping routines for interpreting the user's input, allocating and recovering space, formatting output and handling faults and other messages. It also contains powerful facilities for tabulating and displaying data as well as generalised array and table manipulation capabilities that include many of the features of the APL approach. The use of macros (sets of Prism statements) gives the user a recursive programming language of considerable generality with looping, branching, locally-defined structures and argument-substitution. GLIM-4 is the new release of GLIM, which has become recognised as one of the most useful interactive statistical packages for the analysis of linear models, log-linear models, logit models etc., and is now widely used for the teaching of such skills. It provides a powerful but flexible method for specifying and fitting such models allowing each model to be analysed in a natural, consistent way. GLIM-4 extends the facilities of GLIM-3 in two ways. Firstly the PREDICT directive enables tables of predicted or fitted values to be produced from the fitted model, thus making the output of summary values a simple matter. Secondly a wider range of models is available, including the Inverse Gaussian distribution, certain models used in failure-time and case-control studies and a Package Reports 55 general facility for forming conditional and marginal models. The AOV module analyses all generally balanced designs. These include most commonly-occurring experimental designs, such as randomised blocks, Latin squares, Graeco-Latin squares and other orthogonal designs, as well as designs with balanced confounding, like balanced lattices and balanced incomplete block designs. Many partially-balanced designs can also be handled. Designs may have more than one error term, as with the split-plot design or in repeated-measures designs. The output includes the analysis of variance tables, tables of means with their standard errors, tables of effects, contrasts between effects etc. Full analysis of covariance and treatment of missing values are available. AOV uses the same algorithm as the ANOVA directive of Genstat, and has similar syntax. The new GRAPH module is an easy-to-use, device-independent program for producing high quality graphic output. Standard statistical capabilities include histograms, bar-charts, scatterplots, pie-charts, etc. Other standard features include surface views, contour plots, line-charts, etc. Additionally, low-level features such as the drawing of lines, markers, axes, labelling, colouring, area-infill, etc., permit the creation of user-designed graphics; similarly, more complex features such as the provision of picture segments, input devices, multiple work-stations, generalised text drawing, together with the Prism macro facilities, enable very complex pictures to be displayed and manipulated. To ensure device-portability G R A P H has been written in terms of the new ISO draft graphics standard, GKS. An implementation of GKS, up to version 7.0, will also be available with GRAPH. Prism is due for release towards the end of 1983. Prices will not be available until nearer the release date. • Minitab Minitab is a general purpose, interactive statistics package. The design and syntax was originally based on Omnitab, a computer program developed at The National Bureau of Standards. Minitab is very easy to use: researchers and students who have no experience with computers can make effective use of Minitab with about one hour of instruction. Minitab has been found especially useful for exploring data in the early phases of analysis, for producing plots, tables, and other descriptive summaries, and for regression analysis. Users of Minitab are in most areas of business, industry, government, and education. They include engineers, biologists, chemists, sociologists, psychologists, accountants, statisticians, teachers, and students. As with most statistical software, Minitab is in a state of evolution. Its capabilities in 1982 include: (1) Descriptive statistics such as mean, medians, and maximums.