A Programming Technique for Software in Statistical Analysis
Reinhold Haux · 1982
A new programming technique, called ‘the stepwise reduction of data structures’ is proposed for the development of software, especially in statistics. It will be demonstrated, using some well known examples (PEARSON’s χ2-test, KRUSKAL-WALLIS test), that the programming technique is easy to use. One of the advantages of this technique is that the user can call a program (that deals with statistical analysis) with different kinds of input data structures (and not only with one kind as conventional). In spite of this fact the effort for the programmer is usually not larger than that needed to develop a conventional program. It will be shown that the stepwise reduction of data structures can already be used within application oriented languages, like FORTRAN or PASCAL. It complements very well with WIRTH’s stepwise refinement for the purpose of writing well structured programs. Finally, it will be outlined how to design statistical analysis systems so that this technique can also be used within those systems, for greater flexibility and better detection of errors.