Grade Variation Across Parallel Software Engineering Course Sections

This replication package accompanies the double-anonymous study Grade Variation Across Parallel Software Engineering Course Sections. It contains privacy-reduced derived data, analysis scripts, plotting code, checked reference outputs, figure sources, and anonymous coding materials needed to reproduce the analyses supported by the public release. The released inputs contain derived section-pair gaps, within-offering contrasts, neutral course-group identifiers, rank summaries, and stored simulation summaries. The package supports reproduction of the paper's section-gap summaries, same- versus different-instructor comparisons, course-equal aggregation checks, recurrence analysis, associations with student review measures, course-content profile comparisons, timetable sensitivity analysis, exploratory staffing checks, and manuscript figures. To protect participants and comply with research-ethics and data-use restrictions, the archive does not contain source section-level grade counts, individual review entries or comments, professor or instructor names, author or coder names, original course or section identifiers, detailed schedules, course descriptions, institutional identifiers, or linkage keys. No individual student grades are included. Course groups are represented only by neutral surrogate identifiers, and the coding reliability workbook contains anonymous coding vectors rather than coder identities. Some source-dependent stages cannot be rerun from the public archive, including source extraction, reconstruction of sections, original course-content coding, and fresh simulations that require restricted source counts. These boundaries are documented in REPRODUCIBILITY.md and the analysis-specific documentation. All calculations supported by the released summaries can be rerun using python reproduce.py, which checks generated results against the supplied verification baseline.