Instruction
Group size: 1-3 students | Language: R or Python | Course grade: 30%
Choose one statistical method, implement its main algorithm, test it, and explain what you learned. One well-developed, established method is enough.
Group Members
| Presentation completed | Group | Presentation order | Members | Topic |
|---|---|---|---|---|
| 1 | 3 | Jada Grant | How does learning rate affect the accuracy and efficiency of gradient descent in logistic regression? | |
| Shravya Honne | ||||
| Airianna Jones | ||||
| 2 | 10 | Jennifer Guo | TBA | |
| Jieun Park | ||||
| Duc Pham | ||||
| 3 | 7 | Nicholas White | Genetic Algorithms for Function Optimization: Effects of Mutation on Convergence | |
| 4 | 5 | Shaswat Bhushan Jangam | Variance Reduction Using Control Variates: How Does Correlation Affect Monte Carlo Accuracy? A Comparison with Antithetic Variables and Importance Sampling | |
| Prakash Karki | ||||
| Mangsa Limbu Pheudin | ||||
| 5 | 2 | Katelyn Truong | How do more random draws improve accuracy of Monte Carlo simulation? | |
| Zhe Zhong | ||||
| 6 | 4 | Yasmeen Babagana | The Cost of Adaptation: Model Updating, Validation, and Prediction Risk Under Distribution Shift | |
| Silas Bambil | ||||
| Shudeshna Nandy | ||||
| 7 | 1 | Halimah Akinoso | TBA | |
| Seth Kwarteng | ||||
| Elton Kwashie | ||||
| 8 | 8 | Qiaoting Gao | GradePredictR: An R Package for Student Grade Prediction and Model Evaluation | |
| Zidu Peng | ||||
| Kimia Daliran Saravi | ||||
| 9 | 9 | Hovor Moses Kwame | Generating Random Variables: Inverse Transform and Acceptance Rejection | |
| Juliet Seadey | ||||
| 10 | 6 | Yang Zong | TBA |
Project Overview
What you will produce
Use the same method for all three components. Follow the links for full requirements and rubrics.
| Component | Deliverable | Weight |
|---|---|---|
| Report | A PDF with 8-10 pages of main text explaining the method and results, plus source files | 10% |
| Package | An installable R or Python package with documentation, an example, and tests | 10% |
| Presentation | An oral presentation explaining the method, implementation, and findings | 10% |
Python option: You may build a Python package; see the Python requirements.
Report writing model
The report requires 8-10 pages of main text, including the abstract and main-text figures, tables, and captions. A separate title-only cover page and the end sections are excluded; see the complete page-count rule.
Preview the report template (PDF) to see the report layout and section headings. You may refer to published articles in The R Journal for examples of how to organize and present your work.
Use The R Journal as the writing model for the 8-10 page report: introduce the method and software, demonstrate a reproducible package workflow, and evaluate accuracy, computing time, and limitations. This structure applies to both R and Python projects. See the journal references and report structure and downloadable Quarto starter.
How to get started
- Choose a topic. Use the instructor’s list (TBA), or propose your own method for instructor approval.
- Propose your plan. Send member names, the method, language, core algorithm, and a planned experiment. See approval details.
- Build and compare. Test two settings against a known answer or reference method; explain accuracy, computing time, and limitations.
A few ideas
- Bootstrap intervals: how often do intervals contain the true population mean?
- Gradient descent: how does step size affect convergence when fitting a regression?
- Monte Carlo integration: how do more random draws improve accuracy?
Due on December 13, 2026: report and package. Presentation date, time, and time limit: TBA. All members contribute and present. No slides or other presentation files need to be submitted.