Selecting DataFrame Columns

Table of Contents

• Introduction
• Selecting Columns by Labels
  • Selecting Single Column
  • Selecting Multiple Columns
• Selection by the loc Attribute
  • Selecting a Cell
  • Selecting Single Column
  • Selecting Multiple Columns
• Selection by the iloc Attribute
  • Selecting a Cell
  • Selecting Single Column
  • Selecting Multiple Columns

Introduction

We learned in the previous sections that a DataFrame behaves in many ways like a 2-D NumPy array or a structured array, and in other ways like a Python dictionary of Series objects sharing the same index.

These analogies turn out to be helpful as we study data selection in DataFrames. In this section, we will discuss accessing the columns in a DataFrame by label and by position.

Selecting Columns by Labels

We first import the required libraries and modules.

1import numpy as np
2import pandas as pd
3import random as rd

We now create a dataframe from a dictionary of lists using the random module to generate the required data.

Example

1a=['Tom','Gary','Lois','Wendy','Betty']
2b=rd.choices(range(50,85),k=5)
3c=np.random.uniform(1.6, 1.9, 5).round(2)
4
5df = pd.DataFrame({'Weight':b,
6                   'Height':c}, index=a)
7df

Selecting Single Column

We can access each column using its column label (which is the key of the original dictionary). The output is a Series.

Example

1df['Weight']

Tom      58
Gary     54
Lois     57
Wendy    79
Betty    65
Name: Weight, dtype: int64

Alternatively, we can access each column using the column name as an attribute.

Example

1df.Weight

Tom      58
Gary     54
Lois     57
Wendy    79
Betty    65
Name: Weight, dtype: int64

For example, we create a column with ‘shape’ as its label.

Example

1df1 = pd.DataFrame({'Weight':b,
2                   'shape':c}, index=a)
3df1

In this case, df1.shape will return the shape of the dataframe rather than the ‘shape’ column.

Example

1df1.shape

(5, 2)

The dictionary-like syntax can be used to modify the dataframe object. For example, creating a new column.

Example

1df['Ratio'] = df['Weight']/df['Height']
2df

We can also round off a certain column using the round() method.

Example

1df['Ratio'] = df['Ratio'].round(1)
2df

Selecting Multiple Columns

We can put the labels of all columns we want in a list and feed into the dataframe.

Example

1cols = ['Weight','Ratio']
2df[cols]

Selection by the loc Attribute

The loc attribute is used to access a group of rows and columns by label(s) or a boolean array.

Syntax

1dataframe.loc[row_selection, column_selection]

where row_selection and column_selection refer to a list of row labels and column labels, respectively, i.e. ['label1', 'label2', 'label3']. loc supports the slice notation and therefore accepts a colon(:) to select all rows or columns.

Selecting a Cell

We can select a cell value by specifying both row and column labels.

Example

1df.loc['Tom','Height']

1.65

Selecting Single Column

Select an entire column using its label. The colon (:) in the row position indicates all rows are selected.

Example

1df.loc[:,"Weight"]

Tom      58
Gary     54
Lois     57
Wendy    79
Betty    65
Name: Weight, dtype: int64

The above returns a series object.

1type(df.loc[:,"Weight"])

pandas.core.series.Series

If the column label is in a list, then a DataFrame will be created instead.

1df.loc[:,["Weight"]]

1type(df.loc[:,["Weight"]])

pandas.core.frame.DataFrame

Selecting Multiple Columns

The labels of the columns we wish to select are inside a list. This will always return a DataFrame.

Example

1df.loc[:,['Weight','Ratio']]

We can also select multiple contiguous columns (adjacent to each other with no gap) using the colon : notation.

Example

1df.loc[:,'Weight':'Height']

The usefulness of contiguous column selection may not be apparent in a DataFrame with only a few columns. However, imagine you have a large DataFrame and you need to select over a dozen contiguous columns. In this case, the colon (:) notation will do away with the need to write down the label of every single column included in the selection.

Selection by the iloc Attribute

The iloc attribute is a purely integer-location based indexing for selection by position.

Syntax

1dataframe.iloc[row_selection, column_selection]

where row_selection and column_selection refer to a list of row indices and column indices, respectively.iloc supports the slice notation and therefore accepts a colon(:) to select all rows or columns.

Selecting a Cell

We can select a cell value by specifying both row and column indices.

Example

1df.iloc[0,0]

58

Selecting Single Column

We can select an entire column using its index. The colon (:) in row_selection indicates all rows are selected.

Example

1df.iloc[:,2]

Tom      35.2
Gary     29.3
Lois     33.5
Wendy    45.7
Betty    38.7
Name: Ratio, dtype: float64

The above returns a series object.

1type(df.iloc[:,2])

pandas.core.series.Series

If the column label is in a list, then a DataFrame will be created instead.

Example

1df.iloc[:,[2]]

1type(df.iloc[:,[2]])

pandas.core.frame.DataFrame

Selecting Multiple Columns

Selecting multiple columns (whose indices are inside a list) will always return a DataFrame.

Example

1df.iloc[:,[0,2]]

We can also select multiple contiguous columns (adjacent to each other with no gap) using the colon : notation.

Example

1df.iloc[:,0:2]