Unit 04 · lab

Evidence Check: Build a Data-Cleaning Plan

Imagine a dataset with columns for student_id, team, event_date, score, and notes. Some rows have missing scores, inconsistent dates, duplicate IDs, and team names written several different ways.

Your evidence

Produce a cleaning plan with a table containing:

Problem	How you would detect it	What you would do	What could go wrong

Include at least five checks.

Then use the Robotnix browser data workbench below to run or edit at least three Pandas operations. Record what each operation changes and one piece of information it could accidentally erase or distort.

Break the dataset on purpose

Open the Robot Telemetry Lab. Compare an anomaly, an implausible value, and missing sensor fields. Decide which condition is evidence about the robot and which is primarily evidence about the dataset or sensor pipeline.

Check your reasoning

Cleaning is not deleting everything unusual. Every transformation should have a reason and should preserve an audit trail.

Done when

Every proposed transformation names both the defect it addresses and the information it could accidentally change or remove.

1. Dataset

Fictional robot test runs. No student or personal data.

run	model	sec	battery	°C	done
R-101	Atlas	52	88	22	yes
R-102	Atlas	57	79	23	yes
R-103	Nova	64	74	24	yes
R-104	Nova	71	—	27	no
R-105	Pulse	48	92	21	yes
R-106	Pulse	55	84	25	yes
R-107	Atlas	68	69	29	no
R-107	Atlas	68	69	29	no

2. Quick chart

ViewY

Change the variables before deciding what the data “says.” The same rows can support different questions.

3. Editable Pandas analysis

Runs locally in your browser. First use downloads Pandas and its verified dependencies from Robotnix.

import pandas as pd

rows = [
    {"run_id": "R-101", "model": "Atlas", "duration_s": 52, "battery_pct": 88, "ambient_c": 22, "completed": True},
    {"run_id": "R-102", "model": "Atlas", "duration_s": 57, "battery_pct": 79, "ambient_c": 23, "completed": True},
    {"run_id": "R-103", "model": "Nova", "duration_s": 64, "battery_pct": 74, "ambient_c": 24, "completed": True},
    {"run_id": "R-104", "model": "Nova", "duration_s": 71, "battery_pct": None, "ambient_c": 27, "completed": False},
    {"run_id": "R-105", "model": "Pulse", "duration_s": 48, "battery_pct": 92, "ambient_c": 21, "completed": True},
    {"run_id": "R-106", "model": "Pulse", "duration_s": 55, "battery_pct": 84, "ambient_c": 25, "completed": True},
    {"run_id": "R-107", "model": "Atlas ", "duration_s": 68, "battery_pct": 69, "ambient_c": 29, "completed": False},
    {"run_id": "R-107", "model": "Atlas ", "duration_s": 68, "battery_pct": 69, "ambient_c": 29, "completed": False},
]
df = pd.DataFrame(rows)

print(df)
print("\nMissing values:")
print(df.isna().sum())
print("\nDuplicate run IDs:", df["run_id"].duplicated().sum())

Output

Run the analysis to see evidence.

Ready. Predict before you run.

Runtime boundary: Robotnix serves the core Pyodide runtime and the SHA-256-verified Pandas dependency set. This workbench does not expose pip, micropip, a shell, filesystem projects, cloud credentials, or arbitrary package installation.