How Scratch Helps College Students Learn Programming and Understanding Gender Differences

Experimental Data Sets on Constructionist Scratch Pedagogy in College CS1

These datasets consist of experimental data collected from four cohorts of first-year computer science students from selected polytechnics in north-central Nigeria. From two quasi-experimental studies, these datasets include demographic data and data on student achievement before and after a learning period in their first programming course (CS1). In each study, two cohorts of CS1 students were treated with either constructionist Scratch or conventional programming pedagogy in the experimental and control groups, respectively. By adapting an instrument, we developed two instruments: a student profile questionnaire and an Introductory Programming Achievement Test (IPAT). Students wrote IPAT as a pretest, and after a six-week intervention, with some rearrangement of questions, as a posttest. The Coarsened Exact Matching (CEM) algorithm was used to generate matched cases of experimental and control data used in the analysis. While we tested eight research hypotheses in the original research project, the main research hypothesis was: There is no statistically significant difference in the mean post-test achievement between CS1 students in the constructionist Scratch class and those in the conventional lecture-based class, after controlling for their pretest scores. Findings from paired samples t-test and ANCOVA analyses of the pilot and main study data suggest a consistent outcome: students in the constructionist Scratch outperformed those in the conventional class, although the impart was moderate in both instances. This implies that a constructionist Scratch programming intervention may be more effective with first-year higher education students with no prior experience in programming. These datasets can provide some further insights. For instance, the nature of the programming knowledge and understanding exhibited by the students in the two groups. From their answers to the open-ended questions in the achievement tests, we can explore that. A researcher can replicate the study by employing the same unmatched or matched data, or generating random matched samples from the unmatched dataset. We can also form and test other hypotheses.