Raymond R. Panko
Too much spreadsheet research is unpublishable in high-quality journals due to poor methodology. This is especially a problem for computer science researchers, who often are untrained in behavioral research methodology.
This position paper reflects the author's experiences in reviewing submissions to information systems and computer science journals.
Issues in inspections (audits) of operational spreadsheets, spreadsheet development experiments, and spreadsheet inspection experiments:
- Respect human error research.
- Don't trust. Verify.
- Report time spent.
- Understanding the spreadsheet first.
- Report error seriousness.
- Use new tasks.
- Have adequate task length.
- Go beyond student samples.
- Test prescriptions for safety and effectiveness.
- Go all the way to error reduction.
- Use ample sample sizes.
- Avoid friends and family samples.
- Do rigorous random assignment to conditions.
- Use nonparametric statistics.
- Be generous in presenting statistical results.
- Higher error rates.
- Test for safety and effectiveness.
2014, First Workshop on Software Engineering Methods in Spreadsheets, July