Authors
James A. Hoag
Abstract
The spreadsheet has become a common technology tool and is now a predominant form of end-user programming. Some of the same features that make spreadsheets excellent tools for ad-hoc development can introduce errors into the final product. Although a variety of research has been performed investigating methods to detect errors in spreadsheets, little has been done to investigate initial reasoning errors. The spreadsheet error taxonomy developed by Rajalingham, Chadwick, and Knight (2001) includes categories for reasoning errors that have not yet been investigated. Previous studies have categorized errors in existing spreadsheets, but have not analyzed the source of the error.
This study investigates the reasoning of college students while developing spreadsheets and examines the reasoning associated with errors generated in spreadsheet development. For this study, a phenomenological qualitative design incorporated a think-aloud protocol, interviews, and recordings of spreadsheet development of three purposefully-selected students. Data sources were analyzed to determine their reasoning, the types of errors produced based on the taxonomy, and associations between reasoning and errors.
The findings indicated that students used different types of reasoning in the mathematical phases of spreadsheet development than they do in the spreadsheet implementation phase. As novice spreadsheet developers, the students had significant difficulties translating problems into mathematical representations. Spreadsheet skills and concepts improved with practice through the course, but mathematical representations remained problematic. The students enjoyed using the spreadsheet as a tool for doing mathematical reasoning.
Several themes emerged as the study progressed: Reasoning differences during mathematical and spreadsheet phases of development; using icons for functions affected conceptualization of the functions; copy operations were perceived as "painting" rather than applying a formula to a series; and the effectiveness of the taxonomy for categorizing reasoning errors.
The results suggested modifications to student learning experiences leading to more accurate spreadsheet development: Integrate spreadsheets into mathematics courses; increase education in spreadsheet development; integrate formal design and testing components to the spreadsheet curriculum; include spreadsheet errors in the curriculum.
Sample
Hard-coding of values rather than use of cell references was a common practice and source of qualitative errors in novice spreadsheet developers.
In this example, the formula for Gross Pay was hard-coded for the first entry, while the other cells used a formula.
As a qualitative error, this kind of reasoning had the capability to produce correct results for a specific instance, but provided no flexibility for an application that might be used for a diverse set of inputs.
As evidence that hard-coding was common amongst novice spreadsheet developers, on the final exam, 39% of the spreadsheets contained some form of hard-coding.
Publication
2008, Ph.D thesis, Oregon State University, July
