Marimo Tutorial (III): Data Analysis That Never Breaks

Safe Pandas Analysis with Reactive Guarantees

In the previous articles [1, 2, 3], we compared Marimo Notebook with Jupyter Notebook, how to install Marimo and create a Marimo Notebook, and how to use Marimo UI components like slider. In this tutorial article,

What You’ll Learn

• Use Pandas safely
• Avoid stale variables
• Build reproducible analysis notebooks

Create a Simple Dataset

import pandas as pd

data = {
    "year": [2020, 2021, 2022, 2023],
    "sales": [100, 130, 160, 210],
}
df = pd.DataFrame(data)
df

Add a Filter Control

import marimo as mo

# Create the slider
slider = mo.ui.slider(2020, 2023, value=2021, label="Select Minimum Year")
slider

Reactive Data Filtering

# To get the current value:
min_year_value = slider.value  # <-- use .value here

# Filter DataFrame
filtered_df = df[df["year"] >= min_year_value]
filtered_df

Change the slider:

• Data updates instantly
• No stale DataFrame
• No hidden kernel state

Plot the Results

import matplotlib.pyplot as plt

plt.plot(filtered_df["year"], filtered_df["sales"], marker="o")
plt.xlabel("Year")
plt.ylabel("Sales")
plt.title("Sales Over Time")
plt.show()

The plot always matches the data — guaranteed.

Why This Beats Traditional Notebooks

Common Jupyter problem:

• Filter cell runs before data cell
• Results silently wrong

Marimo solution:

• Dependency graph prevents invalid execution

Key Takeaway

Marimo makes data analysis trustworthy by default.

Final Summary

| Tutorial   | Focus              | What You Gained        |
| ---------- | ------------------ | ---------------------- |
| Tutorial 1 | Core concepts      | Reactive execution     |
| Tutorial 2 | UI & interactivity | App-like notebooks     |
| Tutorial 3 | Data analysis      | Reproducible workflows |