Effects of computer interfaces on computer-based statistical analysis

Abstract

Thirty-one first-year psychology graduate students in a computer applications course completed a set of structured problems, unstructured problems, and data-screening problems in each of two statistical computing environments: a menu-based interface (SPSS for Windows) and a traditional command-based interface (SPSSx). Performance on the menu-based interface was generally superior to performance on the command-based interface. More of the structured problems were completed successfully within the menu-based interface. The menu-based interface also facilitated error identification, was rated as easier to use, and was preferred nearly 4 to 1 over the command-based interface. For the unstructured problems, students identified more relationships in the data set and issued more statistical commands when working with the menu-based interface. These findings are consistent with the interpretation that because the menu-based interface requires fewer mental resources to be dedicated to the mechanics of analysis, more resources are available to devote to higher level problem solving.