SYNOPTIC ABSTRACT
Recently experimental packages with with intelligent interfaces have been developed. We discuss the usefulness of these packages (often called Statistical Expert Systems) during the five recognizable phases of a scientific investigation. We will see that each phase needs its own type of communication. Therefore and for other reasons we conclude that for the expert data analyst there is a need for a good ‘environment’, but, as yet, not for some catch-all expert system for scientific investigation.
Keywords and Phrases: