How to drop rows of Pandas DataFrame whose value in a certain column is NaN

Drop NaN with Different Scopes and Conditions

Selective removal, or how I learned to not kill all the NaNs in sight

Weed out NaN values directly from one or multiple columns:

df = df.dropna(subset=['your_column']) # Single column
df = df.dropna(subset=['col1', 'col2', 'col3']) # Multiple columns
# "We don’t need no nullification; We don’t need no NaN control" 🎵🎸


Dropping NaN in-place; because, who likes re-assignments?

Save the result back into df without additional line of code:

df.dropna(subset=['your_column'], inplace=True)
# 🎶 Hit me NaN one more time! 🎶 Oops..it's gone forever. 😉


Cut-off Thresholds; Data cleaning meets Highjump

Remove rows that don't meet a certain count of non-NaN values (the threshold):

df = df.dropna(thresh=n) # n is the cutoff number of non-NaN values
# Rows failing to jump over our set bar (n) are out of the DataFrame race!


All or nothing, the NaN version

A stricter approach—discard rows that have NaN values in all columns:

df = df.dropna(how='all')
# "All for NaN, and NaN for all!" said no row ever after this round. 😂


The Boolean Mask, not a new Superhero

Create a mask for rows with valid values and apply it:

mask = df['column_of_interest'].notna()
df = df[mask]
# My Boolean Mask brings all the rows to the yard, and NaN’s like: It's better than y'all!


The Whole Shebang: Dropping rows/columns, thresholds and masks

Sweep, don't weep

Get rid of any row that contains at least one NaN value with df.dropna(how='any'):

df = df.dropna(how='any')
# NaN values are like uninvited party guests, they don’t get past this bouncer.


X-rays for your DataFrame

Before deleting, it helps to know where the NaNs are. With isna().any(axis=1) you see the affected rows:

nan_rows = df.isna().any(axis=1)
num_nan_rows = nan_rows.sum()
# It’s like playing ‘Where’s Waldo?’ but with NaN values. 🧐


Null-cypher: Dropping columns with NaNs

What if columns rather than rows are jammed with NaN values? You deal with them by just switching the axis in dropna:

df = df.dropna(axis=1)  # Drops any column with at least one NaN value


Understand NaNs: Sometimes Absence Makes the Data Grow Fonder

NaNs aren't always a problem; they might signify missing information. Sometimes, imputing with a statistical measure (mean, median) or a constant value therefore can be more beneficial. Also, the pattern of NaNs can provide insights on data quality or bias.