Authors
Raymond R. Panko
Abstract
The widespread presence of errors in spreadsheets is now well-established. Quite a few methodological and software approaches have been suggested as ways to reduce spreadsheet errors.
However, these approaches are always tailored to particular types of errors. Are such errors, in fact, widespread? A tool that focuses on rare errors is not very appealing. In other fields of error analysis, especially linguistics, it has proven useful to collect corpuses (systematic samples) of errors.
This paper presents two corpuses of errors seen in spreadsheet experiments. Hopefully, these corpuses will help us assess the claims of spreadsheet reduction approaches and should guide theory creation and testing.
Sample
The two spreadsheet corpuses, both created by students, are:
- The Galumpke corpus. The Galumpke task requires considerable accounting knowledge, so accounting majors were excluded from the analysis because of their specialized domain knowledge. The corpus comes from an experiment using 33 undergraduate students and 49 graduate students. This experiment yielded a corpus of 68 errors.
- The Wall corpus. The Wall task was designed to be rather simple and not to require specialized knowledge. The corpus comes from an experiment using 101 undergraduate students and 49 graduate students. This experiment yielded a corpus of 62 errors.
Publication
2000, Proceedings of the 33rd Hawaii International Conference on System Sciences
