Report
STAT 8670 | Fall 2026
Due on December 13, 2026
Report: 10 points (10% of the course grade)
Group size: 1-3 students
Project total: 30 points (30% of the course grade)
What is the project?
Choose one statistical method, implement it in R or Python, and show how well it works. Your group will produce a reusable package, explain the method and results in a written report, and present the project to the class.
An established method is a suitable topic. The goal is to understand its statistical reasoning, implement its main computational steps, and evaluate the results. Your project should answer:
- What problem does the method solve? Identify the inputs, outputs, and assumptions.
- How does the method work? Explain the algorithm and implement its core steps.
- How well does it work? Use reproducible experiments to examine accuracy, computing time, and limitations.
| Component | What your group submits or delivers | Points |
|---|---|---|
| Report | A PDF with 8-10 pages of main text explaining the method, implementation, and results, plus source files | 10 |
| Package | An installable R or Python package, documentation, a worked example, and tests | 10 |
| Presentation | An oral presentation explaining the project; no presentation files to submit | 10 |
| Total | One coordinated project | 30 |
Each component is worth 10 points and 10% of the course grade: report 10%, package 10%, and presentation 10%. Each page gives the requirements and point-by-point rubric for that component.
Groups and topic approval
Work individually or in a group of 2 or 3 students. Choose one group name and list every member on the report and repository, and introduce them during the oral presentation. Each member should make a meaningful contribution and be able to explain the method and results. Groups submit one shared set of materials and receive a shared project score under the rubrics below.
You may choose a topic from the instructor’s topic list or propose your own statistical method. Self-selected topics require the instructor’s approval. Confirm the scope with the instructor before starting the full implementation.
Instructor’s topic list: TBA
Group and topic proposal deadline: TBA
Report and package: Due on December 13, 2026
Presentation date, time, and time limit: TBA, after groups are finalized
Send the instructor a short proposal with your group members, proposed method, choice of R or Python, core algorithm to implement, and one planned experiment or application. Explain what your group will implement and which existing libraries you plan to use.
Examples of a manageable project
These examples illustrate possible scope; confirm your choice with the instructor.
| Statistical method | What you could implement | What you could investigate |
|---|---|---|
| Bootstrap confidence intervals | Resampling and interval construction for a mean or median | How often do the intervals cover the true value as sample size changes? |
| Numerical optimization | Gradient descent for a regression objective | How do step size and starting value affect convergence, accuracy, and time? |
| Monte Carlo integration | A basic estimator and a variance-reduction method | How does estimation error change as the number of draws increases? |
| A permutation test | Shuffling group labels and calculating a test statistic | Does the test maintain its intended rejection rate under a simulated null model? |
| A regression or dimension-reduction method | The main fitting algorithm and a prediction or transformation function | How do its results compare with a trusted reference implementation? |
For example: “We implement percentile bootstrap confidence intervals for a population mean in Python, compare coverage and interval length at two sample sizes, and explain when the method performs poorly.” A focused question like this is enough to organize all three project components.
Report requirements
Write 8-10 pages of main text, counted using the rule below. Use Quarto, R Markdown, or LaTeX and submit the rendered PDF plus its source files. Use readable 11- or 12-point text, standard margins, and numbered sections. Place long code listings and supplementary output in an appendix or the repository.
The finished report must contain at least 8 and at most 10 pages of main text.
- Count: the abstract, Sections 1-6 (Introduction through Discussion and conclusion), and all figures, tables, captions, equations, and code examples within those sections.
- Do not count: a separate title-only cover page, References, Software availability and reproducibility, Author contributions, and appendices or supplementary material. Put these end sections after the main text; start References on a new page.
- A first page containing the title, author names, and abstract counts as a main-text page. Only a separate page containing the title and author information alone is excluded; a separate cover page is optional.
The 9-page plan below is a suggested allocation, not a different requirement. The short template PDF shows headings and prompts; its number of pages is not the required report length. The full submitted PDF may exceed 10 pages because the excluded material comes afterward.
Writing model: The R Journal
Write the report as a short R Journal-style software article: explain the statistical problem, introduce your implementation, show a usable workflow, and evaluate its behavior. The R Journal author guide emphasizes a broad readership, software design, practical use, and reproducible results. For this course, also explain the statistical formulation and the core algorithm your group implements, as required by the rubric.
This is a course adaptation of the journal’s structure. Both R and Python projects use it. Keep the course’s 8-10 page limit, readable 11- or 12-point text, and PDF-plus-source submission. A course project may implement an established method; journal acceptance, methodological novelty, and CRAN/PyPI publication are not requirements. The existing report, package, and presentation rubrics determine the grades.
Read these references before drafting:
- R Journal author guide: article organization, software reporting, and reproducibility.
- Official rjtools documentation: the journal’s R Markdown authoring workflow. This is an optional tool for R users; the course also accepts Quarto and LaTeX.
- Official article-template showcase: links to source files and a rendered sample article.
- Kenny and Wolock (2025), “SimEngine: A Modular Framework for Statistical Simulations in R”: a published example linking motivation, design choices, a user workflow, and discussion. Use its exposition as a model; your project can remain a small implementation of one method.
Suggested report structure
Use a descriptive title identifying the method or package, and list the group and all members. Start with an abstract of about 150-200 words, covering the problem, implementation, evaluation, and main result. This length is a course writing suggestion. Define technical terms for readers who have not seen your code.
The following is an illustrative 9-page plan, including the abstract and main-text figures and tables. Adjust the allocation within the required 8-10 counted pages; apply the page-count rule above.
| Section | What the reader should learn | Pages |
|---|---|---|
| Abstract | The purpose, implementation, evaluation, and main finding. | 0.5 |
| 1. Introduction | The problem, intended users, existing approaches, and the scope of your package. | 1 |
| 2. Statistical method | The estimand, test, or optimization objective; notation, assumptions, inputs, and outputs. State constraints when relevant. | 1 |
| 3. Algorithm and implementation | Core steps or pseudocode, design choices, tuning/stopping rules, numerical safeguards, and which routines you wrote or reused. | 1.5 |
| 4. Package workflow and example | Installation, a minimal function call, its output, and its statistical interpretation. Use a small example a reader can reproduce. | 1 |
| 5. Experiments and results | Two settings, a known answer or reference method, accuracy and timing comparisons, and an explanatory results figure. | 3 |
| 6. Discussion and conclusion | What worked, computational cost, a limitation or failure case, and one useful extension. | 1 |
For Section 5, give the simulation or data-analysis design, metric definitions, random seeds, replication counts when relevant, and software/hardware details needed to understand the timing results. Use the same experiment conditions for your implementation and the reference. Distinguish correctness checks from performance comparisons. Include at least one reproducible simulation study or data analysis with two settings, as required above.
After the main text, include References, Software availability and reproducibility, and Author contributions. Cite methods, data, packages, and borrowed or adapted code with consistent author-year references. Identify the package repository and the exact release tag or commit used for all results. Give commands for reproducing the main tables and figures. Add a brief contribution table; an individual project should say that the work was completed individually.
Number and caption figures and tables and explain their statistical meaning in the text. A few short, evaluated code examples belong in the workflow section. Keep full source listings, long logs, and additional experiments in the repository or supplementary material. Use the same submitted package version for the report and package assessment.
Preview and download the template
Expand the three-page PDF preview below, open a larger view, or download the PDF. This PDF is rendered directly from the downloadable template below and shows the layout, section headings, and writing prompts. It is a short outline, not an example of the required length; your completed report must contain 8-10 counted pages of main text under the page-count rule.
You may refer to published articles in The R Journal and the examples linked above for software presentation, reproducible examples, figures, and experimental results. Use them as models for organizing and presenting your own work.
Open PDF preview (3 pages)Close PDF preview
Use the page and zoom controls to read the template. Click the heading again to collapse the preview.
Download the STAT 8670 report starter (ZIP). Unzip it to obtain Project_Report_GroupName.qmd and README.md. Replace GroupName and the writing prompts, add your code and citations, and render the matching PDF. The starter uses the R Journal-inspired structure with ordinary Quarto, so it works as a report outline for either an R or a Python package.
Report grading [10]
The first four criteria assess clarity and writing [4]; the last four assess method, experiments, and interpretation [6].
| Criterion | Full-credit evidence | Points |
|---|---|---|
| R1. Organization | Required sections form a logical account of the problem, method, and findings. | 1 |
| R2. Writing and notation | Explanations are readable and technically precise; symbols are defined and used consistently. | 1 |
| R3. Figures and tables | Results are legible, labeled, captioned, and interpreted in the text. | 1 |
| R4. Scientific reporting | The report follows the length and format requirements and includes references and a contribution statement. | 1 |
| R5. Statistical method | The purpose and assumptions are correct [1]; the mathematical formulation is correct and clearly stated [1]. | 2 |
| R6. Algorithm | The computational steps and important implementation choices are explained accurately. | 1 |
| R7. Experiments | Reproducible experiments cover two settings [1]; a meaningful reference or baseline supports the accuracy and timing comparisons [1]. | 2 |
| R8. Interpretation | Conclusions follow from the evidence and discuss computational cost, limitations, and a possible extension. | 1 |
| Total | 10 |
Score each 1-point requirement as 1 for complete and correct evidence, 0.5 for partially correct or incomplete evidence, or 0 for missing or substantially incorrect evidence. The 2-point criteria each contain two 1-point requirements, as shown above. A missing report receives 0/10. Package and presentation marks are assessed separately using their own rubrics.
One member submits the following on behalf of the group to the iCollege project submission area (folder details TBA) by December 13, 2026:
Project_Report_GroupName.pdf.- The matching
.qmd,.Rmd, or.texsource, with supporting files in a ZIP if needed. - The package repository URL and the exact release tag or commit used for the report.
Replace GroupName with your group’s chosen name and use it consistently for all components. Include all member names. Check that the report’s figures and tables can be reproduced from the submitted package and experiment code.
Next: Package requirements and grading | Presentation requirements and grading