After identifying and potentially filling some missing data, another common step in data preparation is removing data that is unnecessary or problematic. You might need to remove entire rows (observations) or entire columns (features) from your DataFrame. For example:

A column might contain redundant information already present in another column.
A column might have too many missing values to be reliably filled.
Specific rows might represent outliers or errors that you want to exclude from analysis.
Rows or columns might simply be irrelevant to the specific question you are trying to answer.

Pandas provides the versatile drop() method to handle both row and column removal.

Removing Rows Using Index Labels

To remove one or more rows, you provide the index label(s) of the rows you want to eliminate to the drop() method. You also typically specify that you are targeting rows using the index argument.

Let's start with a sample DataFrame:

import pandas as pd
import numpy as np

data = {'StudentID': ['S101', 'S102', 'S103', 'S104', 'S105', 'S106'],
        'Math': [85, 92, 78, 88, 95, np.nan],
        'Science': [90, 88, 94, 85, 79, np.nan],
        'History': [76, 81, 85, 70, 88, 60],
        'Notes': ['Good', 'Excellent', np.nan, 'Needs Improvement', 'Good', 'Incomplete']}
df = pd.DataFrame(data)
df.set_index('StudentID', inplace=True) # Set StudentID as the index

print("Original DataFrame:")
print(df)

This produces:

Original DataFrame:
           Math  Science  History              Notes
StudentID
S101       85.0     90.0       76               Good
S102       92.0     88.0       81          Excellent
S103       78.0     94.0       85                NaN
S104       88.0     85.0       70  Needs Improvement
S105       95.0     79.0       88               Good
S106        NaN      NaN       60         Incomplete

Suppose we want to remove the student S106 because their data is incomplete. We can do this using drop():

# Drop a single row by index label
df_dropped_row = df.drop(index='S106')

print("\nDataFrame after dropping row 'S106':")
print(df_dropped_row)

Output:

DataFrame after dropping row 'S106':
           Math  Science  History              Notes
StudentID
S101       85.0     90.0       76               Good
S102       92.0     88.0       81          Excellent
S103       78.0     94.0       85                NaN
S104       88.0     85.0       70  Needs Improvement
S105       95.0     79.0       88               Good

Notice that df.drop() by default returns a new DataFrame with the specified row(s) removed. The original DataFrame df remains unchanged.

To drop multiple rows, pass a list of index labels:

# Drop multiple rows by index labels
df_dropped_rows = df.drop(index=['S103', 'S106'])

print("\nDataFrame after dropping rows 'S103' and 'S106':")
print(df_dropped_rows)

Output:

DataFrame after dropping rows 'S103' and 'S106':
           Math  Science  History              Notes
StudentID
S101       85.0     90.0       76               Good
S102       92.0     88.0       81          Excellent
S104       88.0     85.0       70  Needs Improvement
S105       95.0     79.0       88               Good

Removing Columns

Removing columns is very similar, but instead of providing index labels, you provide column names. The preferred way to specify that you're targeting columns is by using the columns argument.

Let's say the Notes column isn't needed for our analysis. We can remove it:

# Drop a single column by name
df_dropped_col = df.drop(columns='Notes')

print("\nDataFrame after dropping the 'Notes' column:")
print(df_dropped_col)

Output:

DataFrame after dropping the 'Notes' column:
           Math  Science  History
StudentID
S101       85.0     90.0       76
S102       92.0     88.0       81
S103       78.0     94.0       85
S104       88.0     85.0       70
S105       95.0     79.0       88
S106        NaN      NaN       60

Again, this operation returns a new DataFrame. The original df still has the Notes column.

To drop multiple columns, pass a list of column names:

# Drop multiple columns by name
df_dropped_cols = df.drop(columns=['Math', 'Notes'])

print("\nDataFrame after dropping 'Math' and 'Notes' columns:")
print(df_dropped_cols)

Output:

DataFrame after dropping 'Math' and 'Notes' columns:
           Science  History
StudentID
S101          90.0       76
S102          88.0       81
S103          94.0       85
S104          85.0       70
S105          79.0       88
S106           NaN       60

Historically, you might see code using axis=1 to indicate column dropping (e.g., df.drop('Notes', axis=1)). While this works, using the columns argument (e.g., df.drop(columns='Notes')) is generally considered more readable and explicit. Similarly, axis=0 corresponds to dropping rows (the default), but using the index argument is clearer.

Modifying the DataFrame In Place

If you are certain you want to modify the original DataFrame directly without creating a new one, you can use the inplace=True argument.

# Create a copy to modify inplace
df_copy = df.copy()
print("\nOriginal df_copy (first 3 rows):")
print(df_copy.head(3))

# Drop the 'History' column inplace
return_value = df_copy.drop(columns='History', inplace=True)

print("\ndf_copy after dropping 'History' inplace (first 3 rows):")
print(df_copy.head(3))
print(f"\nReturn value when inplace=True: {return_value}")

Output:

Original df_copy (first 3 rows):
           Math  Science  History      Notes
StudentID
S101       85.0     90.0       76       Good
S102       92.0     88.0       81  Excellent
S103       78.0     94.0       85        NaN

df_copy after dropping 'History' inplace (first 3 rows):
           Math  Science      Notes
StudentID
S101       85.0     90.0       Good
S102       92.0     88.0  Excellent
S103       78.0     94.0        NaN

Return value when inplace=True: None

Note that when inplace=True is used, the drop method modifies the DataFrame directly and returns None. Be cautious when using inplace=True, as it permanently alters your data. It's often safer, especially when learning or exploring, to work with the default behavior which returns a modified copy, allowing you to keep track of the original data.

Dropping rows and columns using drop() is a fundamental step in refining your dataset, allowing you to focus on the data most relevant to your analysis.