M2: Preliminary Analysis
The M2 submission in the NHANES worked example shows one illustrative version of this milestone. Use it for structure and depth — your project must be your own.
Purpose
M2 checks that the project can move from proposal to evidence. The milestone should show a documented cohort or analytic sample, a missingness check, a descriptive summary, and a careful interpretation that avoids overclaiming.
Due Timing
Submit by the Week 7 deadline posted in Canvas. This milestone prepares the group for Week 8 dashboard prototyping and Week 9 communication work.
Points and Score Ownership
M2 is worth 5 course points and receives one group score. The checklist below guides the direct 0–5 score; no contribution multiplier or performance-band conversion is used.
The five points are allocated across these broad criteria:
| Criterion | Points |
|---|---|
| Cohort definition and inclusion/exclusion decisions | 1 |
| Missingness check and Table 1-style descriptive evidence | 1 |
| Reproducible source, render, and code-generated outputs | 1 |
| Methods note, interpretation, and claim boundary | 1 |
| AI-use audit and complete, synced submission evidence | 1 |
| Total | 5 |
These are broad judgments, not collections of micro-deductions. A clear, rerunnable descriptive analysis can earn full credit without advanced methods or research-grade polish.
Required Deliverables
Create or update:
Include these files. All six are required — none are optional — but three (table1.csv, missingness-summary.csv, methods-note.md) are small outputs or short notes produced alongside the analysis, not separate projects. Tags follow How To Read Submission Lists.
preliminary-analysis.qmd[required]preliminary-analysis.html[required]table1.csv[required]missingness-summary.csv[required]methods-note.md[required]ai-use-note.md[required]
Committing preliminary-analysis.html here is deliberate: every rendered file explicitly required by an assessment is committed with its source.
What To Submit
Submit your GitHub repository link on Canvas after committing and syncing the files above. In the Canvas comment box, include the latest commit hash and the folder path for the milestone.
Analysis Requirements
Your preliminary analysis should include:
- a clear analytic sample or cohort definition
- a short explanation of inclusion and exclusion decisions
- a missingness check for key variables
- a Table 1-style descriptive summary
- at least one figure or table generated by code
- a methods note that states what the summary can and cannot claim
“Table 1-style” is health-research shorthand for the first table of a paper: one row per characteristic of your analytic sample (for example age, sex, income group), with counts, percentages, and means or medians. Follow the Week 7 worked example for the expected format.
Your group may build on one member’s Assignment 6 work. If you do, add one attribution sentence to methods-note.md, for example: “The preliminary analysis builds on Jordan’s A6 eda-note.qmd.” Export the missingness table by code as missingness-summary.csv; this is an M2-only export and is not an additional A6 submission file.
Reproducibility Expectations
The .qmd file must render in GitHub Codespaces or a documented local environment. Use relative paths, document packages, and avoid manual edits to generated tables or figures.
AI-Use Expectations
AI may help draft code, suggest summaries, or review language.
In your ai-use-note.md, name at least three checks you performed, such as checking row counts, variable definitions, missingness, grouping labels, and whether AI changed the meaning of the result. The Week 3 AI audit checklist and the Week 7 audit categories list more checks to choose from.
Full-Credit Preparation Checklist
A substantively complete submission should show:
- cohort definition is explicit
- missingness is reported
- Table 1-style output includes counts and descriptive statistics
- code-generated output is reproducible
- interpretation is descriptive and non-causal unless justified
- AI-generated help is disclosed and audited
- files are committed, synced, and submitted through Canvas
Definition Of Done
M2 is done when a peer can rerun the analysis, see the same descriptive outputs, and understand the limits of the claims being made.