Assessing the Impact of Game-Based Formative Assessment on Sustained Participation in a Large-Scale MOOC

README

Study context:
This dataset is linked with a study conducted in the SWAYAM MOOC "Introduction to Cyber Security" during the January 2026 cohort. The course had a total enrollment of 9,076 learners.

The study examined game-based formative assessment in a large-scale MOOC. Learners who volunteered for the experiment were assigned to either the Game-Based Learning (GBL) group or the Control group through an automated randomizer. The GBL group received simple game-based formative assessment activities linked with weekly course topics. The Control group received the same or equivalent formative questions in a regular quiz format.

Final analytical sample:
The public anonymized dataset contains 525 learners.

GBL group: 274 learners
Control group: 251 learners

The dataset was prepared after removing duplicate records within each group and overlapping cases across the GBL and Control groups.

Files included in this repository:

1. anonymized_clean_dataset.csv
   This is the main anonymized analytical dataset. It contains the variables required to reproduce the main results reported in the paper.

2. data_dictionary.csv
   This file explains the variables included in the anonymized dataset.

3. dataset_summary.csv
   This file gives a short summary of the final analytical sample and group counts.

4. gbl_activity_mapping.csv
   This file provides the weekly mapping of game-based formative assessment activities used in the GBL group.

5. gbl_mooc_zenodo_dataset.xlsx
   This Excel workbook contains the dataset and related supporting sheets in spreadsheet format.

6. gbl_mooc_results_analysis.py
   This Python script reproduces the main descriptive statistics, group comparisons, tables, figures, and result summaries used in the paper.

7. analysis_results_for_paper.zip
   This ZIP file contains the output tables, result summaries, and figures generated from the analysis script.

8. LICENSE.txt
   This file states the reuse conditions for the dataset.

9. CITATION.txt
   This file provides the suggested citation format for this dataset.

Anonymization:
The dataset has been anonymized before public release. The following direct or sensitive identifiers were removed:

* names
* email addresses
* mobile numbers
* enrollment keys
* ABC IDs
* date of birth
* city
* college name
* roll number
* Google Form timestamps
* user IDs
* open-ended learner comments

Open-ended comments were excluded because they may contain identifying information. The learner IDs in the public dataset are newly assigned anonymous IDs after shuffling the row order.

Important scoring rules:

1. "All learners" analysis:
   Non-attempted assessments were coded as zero. This approach was used to measure realised course performance, because non-attempt is also an important outcome in a MOOC.

2. "Attempters only" analysis:
   Only learners with a recorded score for the relevant assessment were included.

3. active_weeks:
   This variable represents the number of weeks from Week 1 to Week 10 in which the learner had a recorded weekly activity or quiz score.

4. final_quiz_attempted:
   This variable is coded as 1 if a final quiz score was available and 0 if no final quiz score was available.

5. final_quiz_score_zero_coded:
   This variable treats non-attempted final quiz records as zero.

Main tables reproduced by the analysis script:
The Python script reproduces the following tables used in the paper:

* Table 1: Sample cleaning and final analytical sample
* Table 2: Sample characteristics and group balance
* Table 3: Assessment performance by group
* Table 4: Weekly participation and weekly quiz performance
* Table 5: Learner response summary
* Table 6: Exploratory subgroup analysis of participation outcomes
* Table 7: Summary of hypothesis testing

Figures reproduced by the analysis script:
The analysis script also generates figures related to weekly participation and final quiz score distribution among attempters.

How to reproduce the analysis:
To reproduce the analysis, run the Python script:

gbl_mooc_results_analysis.py

The script requires the following input file:

anonymized_clean_dataset.csv

The script will generate the analysis output folder and result files, including manuscript-ready tables, CSV tables, result summaries, and figures.

Software requirements:
The analysis was conducted using Python. The main libraries used were:

* pandas
* NumPy
* SciPy
* Matplotlib
* openpyxl

These libraries are commonly available in Google Colab. If needed, they can be installed using pip.

Data use note:
This dataset is shared for academic transparency and reproducibility. It should be used only for research, teaching, review, and verification purposes. The dataset should not be used to identify or attempt to identify any learner.
For questions about the dataset, please contact the corresponding author through the details provided in the related manuscript.