Assignment 4: Extend the Austrian Hotels Dataset

Create new data and analyze it with Claude Code

Overview

In this assignment, you will use Claude Code to:

Generate a new data table that can be joined to the Austrian Hotels dataset
Join your new table with existing data
Perform an analysis that answers an interesting question
Document your process and reflect on using agentic AI

Deliverables

Submit a single PDF containing:

Your data generation prompt and the resulting table description
Your analysis with results (tables, charts, or statistics)
The key code Claude Code generated (not all of it - just the important parts)
A brief reflection (see Part 4)

Part 1: Design Your Data Table (25%)

Task

Create a new CSV file that joins to the Austrian Hotels dataset.

Your table should:

Have a clear join key (hotel_id, city, city+month+year, etc.)
Contain realistic patterns (not random numbers)
Enable an interesting analysis question

Ideas

Pick an idea for a new dataset that could be joined to the Austrian Hotels data. For example:

Table	Columns	Join Key	Analysis Question
Weather	city, month, year, avg_temp, precipitation, snow_days	city + month + year	Does weather affect occupancy?
Events	event_id, city, event_name, month, year, expected_visitors	city + month + year	How much do events boost hotel prices?

You can create something else. Just make sure it has a clear join key and realistic patterns.

What to submit:

The prompt you gave Claude Code to generate the data
A summary of your generated table (columns, row count, sample rows)
Explanation of the realistic patterns you built in

Part 2: Join and Analyze (40%)

Use Claude Code to join your new table with the existing Austrian Hotels data and answer your analysis question.

Requirements:

Perform at least one join operation
Include descriptive statistics (means, counts, distributions)
Create at least one visualization (chart or graph)
Answer your analysis question with evidence

What to submit:

Your main analysis question
Key results (tables, charts)
Your interpretation - what does the data tell us?

Part 3: Show Your Work (20%)

Include the key pieces of code that Claude Code generated. You don’t need to include everything - focus on:

The data generation script (or key parts of it)
The join and analysis code
Any interesting debugging or iteration

Format: Use code blocks in your PDF. If code is long, include just the important functions.

Part 4: Reflection (15%)

Write 100-150 words reflecting on:

Process: How did you iterate with Claude Code? What did you have to clarify or fix?
Comparison: How did this workflow compare to using chat-based AI (ChatGPT/Claude.ai)? What was easier? Harder?
Trust: Did you verify Claude Code’s output? How? Did you find any mistakes?
Future use: When would you use Claude Code vs chat AI in your own work?

Grading Rubric

Component	Excellent (90-100%)	Good (70-89%)	Needs Work (<70%)
Data Design	Creative, realistic patterns, clear join logic	Reasonable table, some patterns	Random data, unclear joins
Analysis	Clear question, appropriate methods, insightful interpretation	Decent analysis, basic interpretation	Unclear question, minimal analysis
Code Quality	Well-organized, key parts shown	Somewhat organized	Messy or missing
Reflection	Thoughtful insights on process and tools	Basic reflection	Superficial or missing

Tips

Be specific in prompts: “5-star hotels should have 20% higher staff-to-room ratios” is better than “make it realistic”
Verify the data: Ask Claude Code to show summary statistics after generating
Iterate: If the first version isn’t right, refine your prompt
Check joins: Always verify row counts before and after joining
Save your work: Keep the Python scripts Claude Code generates

Submission

Format: PDF
Due: Sunday 23:55 before Week 5
Submit via: Moodle

Questions?

If you have trouble with Claude Code setup, check the install guide or ask in the course forum.

--- title: "Assignment 4: Extend the Austrian Hotels Dataset" subtitle: "Create new data and analyze it with Claude Code" --- ## Overview In this assignment, you will use Claude Code to: 1. **Generate a new data table** that can be joined to the Austrian Hotels dataset 2. **Join your new table** with existing data 3. **Perform an analysis** that answers an interesting question 4. **Document your process** and reflect on using agentic AI ## Deliverables Submit a **single PDF** containing: 1. Your data generation prompt and the resulting table description 2. Your analysis with results (tables, charts, or statistics) 3. The key code Claude Code generated (not all of it - just the important parts) 4. A brief reflection (see Part 4) ## Part 1: Design Your Data Table (25%) ### Task Create a new CSV file that joins to the Austrian Hotels dataset. **Your table should:** - Have a clear **join key** (hotel_id, city, city+month+year, etc.) - Contain **realistic patterns** (not random numbers) - Enable an **interesting analysis question** **Ideas** Pick an idea for a new dataset that could be joined to the Austrian Hotels data. For example: | Table | Columns | Join Key | Analysis Question | |-------|---------|----------|-------------------| | **Weather** | city, month, year, avg_temp, precipitation, snow_days | city + month + year | Does weather affect occupancy? | | **Events** | event_id, city, event_name, month, year, expected_visitors | city + month + year | How much do events boost hotel prices? | You can create something else. Just make sure it has a clear join key and realistic patterns. ### What to submit: - The **prompt** you gave Claude Code to generate the data - A **summary** of your generated table (columns, row count, sample rows) - Explanation of the **realistic patterns** you built in ## Part 2: Join and Analyze (40%) Use Claude Code to join your new table with the existing Austrian Hotels data and answer your analysis question. ### Requirements: - Perform at least **one join** operation - Include **descriptive statistics** (means, counts, distributions) - Create at least **one visualization** (chart or graph) - Answer your analysis question with evidence ### What to submit: - Your main **analysis question** - Key **results** (tables, charts) - Your **interpretation** - what does the data tell us? ## Part 3: Show Your Work (20%) Include the key pieces of code that Claude Code generated. You don't need to include everything - focus on: - The data generation script (or key parts of it) - The join and analysis code - Any interesting debugging or iteration **Format:** Use code blocks in your PDF. If code is long, include just the important functions. ## Part 4: Reflection (15%) Write 100-150 words reflecting on: 1. **Process:** How did you iterate with Claude Code? What did you have to clarify or fix? 2. **Comparison:** How did this workflow compare to using chat-based AI (ChatGPT/Claude.ai)? What was easier? Harder? 3. **Trust:** Did you verify Claude Code's output? How? Did you find any mistakes? 4. **Future use:** When would you use Claude Code vs chat AI in your own work? ## Grading Rubric | Component | Excellent (90-100%) | Good (70-89%) | Needs Work (<70%) | |-----------|---------------------|---------------|-------------------| | **Data Design** | Creative, realistic patterns, clear join logic | Reasonable table, some patterns | Random data, unclear joins | | **Analysis** | Clear question, appropriate methods, insightful interpretation | Decent analysis, basic interpretation | Unclear question, minimal analysis | | **Code Quality** | Well-organized, key parts shown | Somewhat organized | Messy or missing | | **Reflection** | Thoughtful insights on process and tools | Basic reflection | Superficial or missing | ## Tips - **Be specific in prompts:** "5-star hotels should have 20% higher staff-to-room ratios" is better than "make it realistic" - **Verify the data:** Ask Claude Code to show summary statistics after generating - **Iterate:** If the first version isn't right, refine your prompt - **Check joins:** Always verify row counts before and after joining - **Save your work:** Keep the Python scripts Claude Code generates ## Submission - **Format:** PDF - **Due:** Sunday 23:55 before Week 5 - **Submit via:** Moodle ## Questions? If you have trouble with Claude Code setup, check the [install guide](../da-knowledge/install-cli.html#claude-code) or ask in the course forum.