Coding Colab

πŸ™Œ Five Claims About a Dataset You’ve Never Seen

This morning you learned the exploration ritual on a dataset we walked through together. Now you’ll use it on one nobody has explained to you.

Work in pairs, in one shared notebook, taking turns at the keyboard. Swap every time you finish a numbered task. The person not typing reads the output aloud and says what they think it means before you move on.

You have 45 minutes.

The data

US National Park Service visitor records:

https://eds-217-essential-python.github.io/data/national_parks.csv

That’s all you’re getting. No data dictionary, no column descriptions. Figuring out what’s in there is the exercise.

Setup

Create a notebook named Colab_2C_Unfamiliar_Data.ipynb, with both partners’ names in the title cell, then read the file in.

Your goal

By the end of the session, produce five factual claims about this dataset, each written in a markdown cell, each immediately preceded by the code cell that proves it.

A claim is factual if a skeptical reader could check it against your output. Compare:

❌ β€œThe dataset has a lot of parks in it.”

βœ… β€œThe dataset contains 4,743 rows describing units of type National Park, out of 21,560 rows total.”

At least one of your five claims must be about a problem with the data.

Working through it

Don’t try to write claims immediately. Explore first, then decide what’s worth claiming.

  1. Size it up. How many rows and columns? What are the column names?

  2. Check the types. Run .dtypes. At least one column has a type you would not predict from its name. Find it, and then work out why by looking at what’s actually in that column. This is the single most important thing in this dataset.

  3. Run the health check. Which columns have missing values, and how many?

  4. Count the categories. Use .value_counts() on at least three text columns. Which columns are informative? Is any of them close to a constant?

  5. Summarize the numbers. Run .describe(). One of the numeric summaries is distorted. Say which, and connect it to what you found in step 2.

  6. Make a working subset. Build a list of the columns you’d actually keep, select them into a smaller DataFrame, and rename at least one column with a dictionary.

  7. Write your five claims. Each one: a code cell, then a markdown cell stating the claim in a full sentence with the specific numbers in it.

If you get stuck on step 2

Ask the question a different way. If a column of years isn’t stored as a number, then something in it isn’t a year. .value_counts() will show you what.

If you finish early

Write a sixth claim about something you’d need to know before using this data to answer β€œwhich national park is the most visited?” You do not need to answer that question. We build the tools for it over the next three days. Just say what would have to be true first.

Wrap-up

We’ll hear two or three pairs read out their problem claim. Be ready to say how you found it, not just what it was.

Resources