Altcademy - a Forbes magazine logo Best Coding Bootcamp 2023

How to select certain columns in Pandas

Understanding DataFrames and Columns

Before we delve into the specifics of selecting columns in Pandas, let's first understand what DataFrames and columns are in the context of Pandas. Think of a DataFrame as a big table, much like one you would find in a spreadsheet program like Microsoft Excel. This table is made up of rows and columns, where each row represents an individual record, and each column represents a particular attribute or feature of that record.

For example, if you have a table of fruits, each row could be a different fruit, and the columns could represent attributes such as name, color, and price.

Accessing DataFrame Columns

To select columns in Pandas, you first need to have a DataFrame to work with. Let's create a simple DataFrame to use throughout our examples:

import pandas as pd

# Create a simple DataFrame
data = {
  'Name': ['Apple', 'Banana', 'Cherry', 'Date'],
  'Color': ['Red', 'Yellow', 'Red', 'Brown'],
  'Price': [1.2, 0.5, 1.5, 1.0]

df = pd.DataFrame(data)

Now, df is our DataFrame, and it contains three columns: 'Name', 'Color', and 'Price'.

Using Square Brackets

The most straightforward way to select a single column is by using square brackets [] with the column name as a string. This will return a Pandas Series, which you can think of as a single column from your DataFrame.

# Selecting the 'Name' column
names = df['Name']

If you want to select multiple columns, you can pass a list of column names inside the square brackets.

# Selecting 'Name' and 'Price' columns
name_and_price = df[['Name', 'Price']]

Dot Notation

Another way to access a column is by using dot notation. This is as simple as accessing an attribute of an object in Python.

# Selecting the 'Color' column using dot notation
colors = df.Color

However, dot notation has its limitations. It won't work if the column names have spaces or if they conflict with DataFrame methods. For example, df.mean would not refer to a column named 'mean'; it would call the method to calculate the mean of each numeric column.

Advanced Column Selection

Sometimes, you might want to select columns based on certain conditions or patterns. Here's where more advanced techniques come into play.

Using loc and iloc

The loc and iloc indexers allow for more advanced selection capabilities. loc is label-based, which means you use the actual labels of your index or column to select data, while iloc is position-based, meaning you use the integer index to select data.

Selecting Columns with loc

To select columns with loc, you can do the following:

# Selecting 'Name' and 'Color' columns using loc
name_and_color = df.loc[:, ['Name', 'Color']]

Here, the colon : before the comma means "select all rows," and the list after the comma specifies which columns to select.

Selecting Columns with iloc

With iloc, you select columns by their integer location:

# Selecting the first and third columns using iloc
first_and_third_column = df.iloc[:, [0, 2]]

Remember that Python uses zero-based indexing, so the first column is at index 0.

Conditional Selection

What if you want to select columns based on a condition? For instance, you might want to select all columns where the average value is greater than a certain threshold. Here's how you can do it:

# Select columns where the mean is greater than 1
mean_greater_than_one = df.loc[:, df.mean() > 1]

In this example, df.mean() > 1 creates a boolean mask, where only columns satisfying the condition have True values. loc then uses this mask to select the appropriate columns.

Intuition and Analogies

Let's use an analogy to reinforce what we've learned. Imagine a bookshelf representing your DataFrame. Each book on the shelf is a column. Using square brackets to select a column is like picking a specific book by its title. Dot notation is like calling a book by a nickname that you've given it, which only works if the nickname is unique and doesn't conflict with other objects on the bookshelf.

Advanced selection with loc and iloc is like asking a librarian (who knows the layout by heart) for books: "I want all books (:) from the 'Fiction' and 'Mystery' categories (the list after the comma)." The librarian understands your request and hands you the books.

Conditional selection is like telling the librarian, "I want all books that have more than 300 pages." The librarian then checks each book and only gives you the ones that meet your criteria.


Selecting columns in Pandas is a fundamental skill that can be compared to picking books from a shelf or filtering through items based on specific criteria. Whether you're using simple square brackets for quick access, loc and iloc for precise control, or conditional statements to filter your selection, the process is about knowing what you need and understanding how to ask for it in the language of Pandas.

As you become more comfortable with these techniques, you'll find that manipulating and analyzing data becomes much more intuitive. And just like a well-organized bookshelf, a well-managed DataFrame is a joy to work with. So go ahead, experiment with the examples provided, and start crafting your own data stories with the power of Pandas at your fingertips.