COGS108 Final Project

Overview

• Goal: Test how well the UCSD student “10% rule” predicts the chance of moving from the waitlist into COGS classes (FA22–SP24).
• Question: Is ~10% of total seats a reliable estimate of the number of students who will enroll from the waitlist, and how does this vary by course and quarter?
• Scope: Six quarters of Cognitive Science courses (undergraduate only), excluding cancelled classes, graduate sections, and independent studies.

Repository Contents

• FinalProject_Group027-FA24.ipynb – final report notebook with narrative, analysis, figures, and discussion.
• csv_data_cleaning.ipynb – cleaning pipeline from the raw UCSD Historical Enrollment Data to the per-quarter cleaned CSVs.
• csv_exploratory_analysis.ipynb / csv_exploratory_analysis.py – exploratory analysis and utility functions used during development.
• Cleaned_data/ – cleaned per-quarter snapshots (fa22_simple.csv, wi23_simple.csv, …) used in the final analysis.
• time_series.csv – full time-series scrape (~1M rows) with enrollment, waitlist, and seat-change history.
• winter_summary.csv – summary table used for intermediate validation.

Data Sources

• Original raw data: UCSD Historical Enrollment Data repository (WebReg scrape).
• Cleaned dataset variables (common across per-quarter CSVs):
  • sec_id, sec_code, prof, course_id
  • total (max seats), ten_percent (0.1 × total)
  • from_waitlist/off_waitlist (students moved from waitlist)
  • size_change (seat count changes during the term)
• Time-series file adds timestamps, availability, waitlist counts, and enrollment counts across scrapes.

Methods

• Filtered to COGS courses; removed seminars/independent study (99, 190+), graduate sections, and cancelled classes.
• Computed ten_percent threshold per class; compared to actual waitlist movement (from_waitlist/off_waitlist).
• Statistical tests on nonparametric data:
  • Spearman correlation between class size and waitlist enrollments (p << 0.05, positive correlation).
  • Wilcoxon signed-rank test comparing the 10% threshold vs. observed waitlist enrollments (failed to reject null at α = 0.05).
• Examined seat-count changes by quarter; ~24% of sections changed capacity at least once (mostly single adjustments).

Key Findings

• Across six quarters, 49.2% of COGS classes met or exceeded the 10% rule threshold.
• Larger classes tended to pull more students from the waitlist (positive but weak relationship).
• No statistically significant difference between the 10% rule expectation and observed waitlist enrollments under Wilcoxon test → retain null hypothesis.
• Capacity changes are common enough to matter, but extreme multi-change cases are rare; classes with arbitrary size changes were excluded from the main analysis.

Limitations and Notes

• Analysis covers only COGS undergraduate offerings from FA22–SP24; results may not generalize to other departments or time spans.
• Independence and normality assumptions are violated in some models; linear fits were used descriptively.
• Seat-count changes and special-topic courses can distort the 10% heuristic; these were filtered when possible, but edge cases remain.

Contributors

Kris Chen, Beijie Cheng, Miles Davis, Alex Woods, Andrea Ruiz D'Argence