Week 3: From Raw Data to Report

Module: Use | Deliverable: Directed report with AI use log (due before Week 4 begins) | See assignment →

Learning objectives

By the end of this week, you should be able to:

Distinguish a vibe report from a directed report
Explore the provided dataset using AI + pandas, identifying data quality issues that require code - not just eyeballing
Build a verified data dictionary documenting variables, missingness patterns, and cleaning decisions
Clean data at scale with assertions after every step
Conduct a formal statistical test and report results with effect sizes and confidence intervals
Write a directed report with a methods section, results, interpretation, and limitations

Sessions

Session 1 · Class 5

From messy data to understanding (at scale)

We enter the Use module with a core distinction: vibe reports vs. directed reports. You'll work with a business dataset that's large enough that you can't eyeball every row. Use AI + pandas to discover data quality issues, compare what AI finds to what your code reveals, and build a verified data dictionary.

Vibe vs. directed reports Data-to-report pipeline AI vs. code discovery at scale Data dictionary workshop

Session 2 · Class 6

Cleaning, analysis, and statistical evidence

You'll write your own cleaning pipeline (the notebook gives requirements, not step-by-step code), learn how to include formal statistical tests in a directed report (t-test, ANOVA, chi-square, confidence intervals), and start building the report with a proper methods section.

Python: cleaning at scale Statistical evidence in reports Work session: analysis and drafting

Prepare before class

Before Session 1:

Skim Békés, Data Analysis with AI, Week 2: From Raw Data to Report - focus on the discover > document > clean > report pipeline
Make sure your Python environment and AI tools are working (see the setup guide if needed)
Review the Python for analysts guide for pandas reference

Before Session 2:

Review your Session 1 data quality findings
Choose which analysis question you want to pursue
Brush up on basic hypothesis testing if needed

Key concepts

Concept	What it means
Vibe report	An unfocused “analysis” with no clear question - what AI produces when you say “analyze this data”
Directed report	Analysis driven by a specific question, with a methods section, verified evidence, formal statistical tests, and limitations
Data dictionary	Documentation of every variable: name, type, description, missingness pattern, issues, and cleaning decisions
Methods section	How you cleaned the data, what you computed, what test you used/ why, sample size, and exclusions
Effect size	How large the difference or relationship is, not just whether it’s statistically significant

Readings and resources

Békés, Gábor. Data Analysis with AI, Week 2: From Raw Data to Report - the discover/document/clean/report pipeline this week is built on, including the data dictionary and vibe-vs-directed report distinction

Deliverable

Directed Report with AI Use Log

4 components | 30 points | Due before Week 4's first class | Submit via Canvas

README + data dictionary, reproducible notebook with assertions, a 3–4 page directed report with methods section and formal statistical test, and an AI use log.

Full prompt and rubric →

Looking ahead

Next week: Debugging AI’s Work - you’ll learn to find bugs in AI-generated code, quantify their impact, and build an automated test function. The central lesson is that output that looks right isn’t necessarily right.