Installing Data Science Libraries (pip & conda)

Data science libraries you’ll use often

A common starter stack for data analytics includes:

• NumPy: numerical computing
• Pandas: data manipulation
• Matplotlib: plotting foundation
• Seaborn: statistical visualization
• Plotly: interactive charts
• Jupyter: notebooks
• SciPy (optional early): scientific utilities
• scikit-learn (later): ML utilities

pip vs conda (how to choose)

Use conda when

• You’re using Anaconda/Miniconda
• You want fewer build/compile issues
• You need compiled dependencies (common in data science)

Use pip when

• You installed CPython from python.org
• You’re inside a venv
• A package isn’t available via conda

Tip

Simple rule

• If you created your environment with conda, install with conda first.
• If you created your environment with venv, install with pip.

Installing with conda

Step 1: Create and activate an environment

command

conda create -n analytics python=3.12

command

conda activate analytics

Step 2: Install the core stack

command

conda install numpy pandas matplotlib seaborn jupyter

Step 3: Install Plotly

Plotly is often available via conda, but some users prefer pip. Try conda first:

command

conda install plotly

If not available in your channels, use pip:

command

pip install plotly

Installing with pip (venv)

Step 1: Create and activate

command

python -m venv .venv

command

source .venv/bin/activate

Step 2: Install packages

command

pip install numpy pandas matplotlib seaborn plotly jupyter

Verifying installs in Python

After installing packages, verify them in a Python session or notebook:

verify

import numpy as np
import pandas as pd
import matplotlib
import seaborn as sns
import plotly
 
print("NumPy:", np.__version__)
print("Pandas:", pd.__version__)
print("Matplotlib:", matplotlib.__version__)
print("Seaborn:", sns.__version__)
print("Plotly:", plotly.__version__)

Installing Jupyter kernel for your environment

Sometimes Jupyter is installed globally but you want the kernel to point at your environment.

Install ipykernel:

command

pip install ipykernel

Register the kernel:

command

python -m ipykernel install --user --name analytics --display-name "Python (analytics)"

Now your environment appears in Jupyter kernel selection.

Reproducibility: pinning versions

For long projects, pin versions so your notebook still runs months later.

pip: requirements.txt

text

numpy==2.1.0
pandas==2.2.3
matplotlib==3.9.2
seaborn==0.13.2
plotly==5.24.1
jupyter==1.1.1

conda: environment.yml

yaml

name: analytics
channels:
  - conda-forge
dependencies:
  - python=3.12
  - numpy
  - pandas
  - matplotlib
  - seaborn
  - plotly
  - jupyter

Caution

Pinned versions are examples Exact versions depend on what’s current and compatible in your environment. Use pip freeze or conda env export to capture your real versions.

Visualize it

Use this decision path whenever you’re not sure which installer to reach for.

diagram pip vs conda decision mermaid

  flowchart TD
  A["Need to install a package"] --> B{"Environment created with conda?"}
  B -->|"Yes"| C["Try: conda install package"]
  C --> D{"Available on conda-forge?"}
  D -->|"Yes"| E["Installed via conda"]
  D -->|"No"| F["Fallback: pip install package"]
  B -->|"No (venv/plain Python)"| G["pip install package"]

Which installer to use depending on how your environment was created

From the book

McKinney’s rule of thumb: conda install should always be preferred when you’re using Miniconda/Anaconda, but some packages simply aren’t on conda channels — in that case, fall back to pip install package_name. The gotcha he flags is mixing update tools: avoid upgrading a conda-installed package with pip (and vice versa), since the two package managers don’t track each other’s metadata and can leave your environment in a broken, inconsistent state.

Common errors and fixes

Error: ModuleNotFoundError: No module named 'pandas'

• You installed in one environment but are running Python from another.
• Solution: activate the correct environment and reinstall.

Error: Jupyter doesn’t show the right kernel

• Install and register ipykernel as shown above.

Error: pip installs but import fails

• Check you’re using the intended pip:
  • In a terminal inside the environment, run which pip (macOS/Linux)
  • Or where pip (Windows)

Next

Phase 1 is complete. Next we’ll start Phase 2: Numerical Computing (NumPy) with an Introduction to NumPy.

🧪 Try It Yourself

Exercise 1 – Choose pip or conda

pch.exerciseTag pch.exerciseTitle

Exercise 2 – Build a requirements.txt From a Dict

pch.exerciseTag pch.exerciseTitle

Exercise 3 – Verify Imports Like a Setup Script

pch.exerciseTag pch.exerciseTitle