How do I do a complete first look at a table and write it down?

Since the first module you've been jotting things down: three repeated trips, eight missing values that go together, one capitalized label, three trips that last almost a day. That list is the log, and it's what a first look really delivers. Charts help you find things; the log is what's left when you close the table.

A good entry has four parts:

What you saw, in the column's own terms: 73 trips without anio_nacimiento.
How many rows, so whoever reads it knows whether it matters.
What you suspect, marked as a suspicion: occasional users probably aren't asked.
What to ask, and whom: confirm with whoever runs Rueda.

The log doesn't fix anything. It does three things: it keeps you from repeating the same search tomorrow, it tells whoever cleans the table what to do, and it lets your report state its limits ("age describes subscribers only") instead of hiding them.

Now you run the complete first look on your own, the way you would with a table someone emailed you. Instead of a single cell you have a small project: a file with the trips, a Python script with the first three sections done, and an empty bitacora.md file for your log. The terminal is already in the project's , which is why the script can open viajes.csv by its name alone. You type uv run primer_vistazo.py and the terminal answers with what the script prints.

Your turn

primer_vistazo.py already prints sections 1 to 3. Complete 4, 5 and 6 so the script prints, in this order:

in 4, the median and the maximum of duracion_min, each on its own line, as print("median:", ...) and print("max:", ...);
in 5, the counts of tipo_usuario with print(viajes["tipo_usuario"].value_counts().to_string());
in 6, print("under 1 min:", ...) and print("over 600 min:", ...), each with the number of trips.

Run it in the terminal with uv run primer_vistazo.py. Then fill in bitacora.md with what you found: one line per finding, with how many rows and what you suspect. You're done when the output shows all six sections and your log has at least one entry in each.

Files

# First look at a new table: Rueda's trips.

# Sections 1 to 3 are already done. Complete 4, 5 and 6.

import pandas as pd

viajes = pd.read_csv("viajes.csv", parse_dates=["inicio"])

print("== 1. One row ==")

print("rows:", len(viajes))

print("distinct trips:", viajes["viaje_id"].nunique())

print("== 2. Shape and types ==")

print("columns:", viajes.shape[1])

print("from:", viajes["inicio"].min())

print("to:", viajes["inicio"].max())

print("== 3. What's missing ==")

print(viajes.isna().sum().to_string())

print("== 4. Each column ==")

# TODO: print the median and the maximum of duracion_min

print("== 5. Categories ==")

# TODO: print the counts of tipo_usuario

print("== 6. Extremes ==")

# TODO: count the trips under 1 minute and the ones over 600

Python 3.12· runs with your account

Compare your output with what you already knew from the earlier modules: 311 rows and 308 distinct trips, 8 missing values in three columns and 73 in birth year, a median of 19.1 minutes, three user-type labels, 6 trips under one minute and 3 over 600. If a number doesn't match, your script is counting something different from what you think: read it line by line before moving on. If an error stops you, paste the message for your Maestro.

The log is the part no script prints. Read your file as if you were someone who has never seen the table: would they know which columns to handle carefully and what to ask? An entry like there are missing values is no use to them; 8 trips without duration, destination or distance, six of them on March 15 after 10 p.m. is. You can ask your Maestro to review your entries and tell you which ones stand on their own.

Save the log next to the data and note at the top which file and which dates it covers. A log without the table it refers to is of little use, and a new table with the same shape (April's trips, say) goes faster if you have what you found in March at hand: you look for the same problems first and note whether they're still there. If you download the project, you take the script, the table and the log together.

One last thing: the order in which someone reads your findings is not the order in which you found them. Whoever asked you for a number cares first about what changes that number. Practice that with the question below.

Check yourself: Your boss needs the typical length of a Rueda trip for a report due today. With your log in hand, what do you tell her first? Answer in two or three sentences.

Answer it yourself