Academic journal article College Student Journal

To Share or Not to Share: Unique Data Sets Facilitate Performance in a Psychology Statistics Course

Academic journal article College Student Journal

To Share or Not to Share: Unique Data Sets Facilitate Performance in a Psychology Statistics Course

Article excerpt

An experiment examined the relation between students analyzing unique versus shared data sets for a psychology statistics homework assignment and their subsequent performance on an exam. Male and female undergraduates first analyzed either a common or unique data set for an assignment on repeated measures analysis of variance. Several weeks later, participants completed a regularly scheduled exam covering the target material. Students who previously had analyzed unique data sets scored significantly higher on both computational and conceptual exam items than did students who had analyzed shared data sets. Thus, analyzing unique data sets appeared to promote learning of statistical principles and procedures. Benefits and limitations of using the unique data set strategy are discussed.


Maximizing students' understanding of statistical procedures while simultaneously meeting the demands of preparation and grading time can be a daunting task, even for experienced instructors. In an effort to assuage the traditionally high workload, many instructors use artificially created data sets for illustrative and homework purposes. Even though some instructors spend additional time creating individual data sets for each student, most simply provide identical data sets to all of the students. Although using unique data sets is more work for the instructor, the approach has been held to be more advantageous to the students for learning statistical concepts (Cake & Hostetter, 1986, 1992; Walsh, 1992). The unique, as opposed to shared, data set approach likely provides an incentive for students to engage in a deeper level of information processing regarding the underlying principles of the methods, thereby enhancing learning.

To begin, analyzing unique data sets should encourage students to become more directly engaged in the learning process. Although students still might consult one another on the process of analyzing data, accountability for analyzing a specific data set likely bolsters active participation at the individual level. Craik and Lockhart (1972) suggest that information processed deeply, or more completely, has a better chance of being retained in memory than information processed only at the surface level, and being actively involved in the learning process appears to promote deeper processing (e.g., Craik & Tulving, 1975). Thus, to the extent that analyzing unique data sets encourages more active learning and deeper processing than analyzing shared data sets, learning should be facilitated.

However, for reasons such as constraints on grading time or limited access to computer programs, many instructors provide only one data set for their students to analyze. Unfortunately, presenting common data sets for homework assignments and other computational exercises may tempt students to take a more passive approach to learning, engaging only in surface level processing of the relevant information. For example, students may focus less on understanding the concepts than on obtaining the same values that their fellow classmates obtain. In addition, they may be more inclined to work in collaboration rather than independently (Cake & Hostetter, 1986, 1992). Thus, saving valuable time by using common data sets may result in an inappropriate learning focus, ultimately hampering rather than helping students' learning. Theoretically, then, students analyzing unique rather than common data sets should work more independently and focus (appropriately) on how to analyze and interpret the data (Cake & Hostetter, 1986, 1992). Both of these factors should facilitate a deeper level of information processing and subsequent learning.

Fortunately, artificial data sets are relatively easy to create and frequently are ideal for illustrating particular concepts. For example, specific data sets may be constructed to demonstrate the consequences of violating the assumptions of analysis of variance. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed


An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.