According to Professor John Neldor, one of the world’s leading experts on statistical modelling, existing statistics packages give the user some help on how to do things, but almost no help on what to do. Oxford, UK-based Numerical Algorithims Group Ltd, NAG, has launched an expert system for statistical analysis, which it claims, redresses this imbalance by encapsulating statistical expertise in a form directly accessible to the user. Glimpse – Generalised Linear Interactive Modelling + Prolog + Statistical Expertise – is the culmination of a five-year UK Government Alvey project by the Group in collaboration with Imperial College, London. It is described as a user friendly, knowledge-based front end to the well-established Glim 3.77 generalised linear modelling – statistics package, now estimated to have over 1,400 users worldwide. Running Glim as a sub process and executing a Prolog interpreter, Glimpse sits on top of an abstract statistician, a body of expert statistical knowledge which can be called upon by the user for guidance and help with problems – or to provide a complete model building strategy, if required. It has a task-based, menu-driven command language, freeing the user from detailed knowledge of the Glim language, an on-line manual and an interactive question and answer interface environment. Glimpse is presently available on Sun workstations running Unix BSD 4.2, priced at UKP7,560 a year at a commercial site – UKP2,520 for academic users. It is particularly appropriate for research in the areas of medicine, agriculture, insurance, econometrics and meteorology according to Professor Nelder. He went on to say that if the statistical methods and models used by Glimpse had been employed by NASA scientists prior to the Challenger space shuttle launch, they would have revealed a 13% probability of catastrophic failure which would have been great enough to have prevented the fated mission from taking off.
