| New developments in statistical methodology coupled with advances in computer hardware have resulted in a significant increase in the number and variety of statistical software packages available to the user. These packages are easy to run and provide enormous computing power, but they do not provide guidance in preparing the data, selecting the analysis methods, and interpreting the computed results. This can easily lead to misinterpretation of the results, especially if the user has a limited knowledge of statistics. To alleviate these problems, statistical consultants provide guidance in using statistics and these packages. However, consultants are scarce in comparison with the number of users.;Based on a study of statistical consulting processes, a theoretical model for a generic process was established. This model provided a framework for constructing an expert system called STATCAN (STATistical Consulting on ANova). The targeted users of this system are industrial engineers and scientists who may be either frequent or casual users of statistics. The system's problem domain is one-way and two-way analysis of variance with or without repeated measures.;Verbal protocol (audio- and videotaping of the expert's ruminations) was used to elicit strategies used by the statistical consultants. Strategies were also extracted by reviewing videotapes of statistical consulting sessions. These strategies, along with the textbook information and knowledge acquired by the investigators through years of statistical consulting, formed the knowledge bases for the expert system.;STATCAN identifies and structures the problem, selects a suitable analysis of variance method, recommends an appropriate software package, and gives a summary report, using an example to illustrate the setup of the package, the use of diagnostic tools, and the interpretation of the results. Graphic displays are included in the summary report where appropriate. Graphical representation of the statistical concepts and terms is also provided. |
