Standard deviation is the square root of the variance, expressed in the same units as the original data, which makes it more directly interpretable than variance alone — a standard deviation of 5 for data measured in meters means typically about 5 meters from the mean. By default, NumPy computes the population standard deviation, dividing by N, but passing ddof=1 computes the sample standard deviation instead, dividing by N-1, the version generally preferred when your data is a sample used to estimate a larger population's variability.

1Understanding np.std()

💡

Pass ddof=1 when your array is a sample meant to estimate a larger population's standard deviation, not the entire population itself — NumPy's default of ddof=0 computes the population version, which several other tools, like pandas, default differently on.

editor.html

import numpy as np

arr = np.array([2, 4, 4, 4, 5, 5, 7, 9])
print(np.std(arr))

localhost:3000

2Practical Example

Here is a real-world application of np.std() showing how it is used in production NumPy code.

editor.html

import numpy as np

sample = np.array([2, 4, 4, 4, 5, 5, 7, 9])
print(np.std(sample, ddof=0))
print(np.std(sample, ddof=1))

localhost:3000

3Best Practices

Follow these guidelines when working with np.std():

1. Use ddof=1 for sample-based statistical estimates, matching the convention many statistics courses and other tools default to

2. Combine std() with mean() to describe a distribution's center and spread together, rather than reporting either alone

3. Use np.nanstd() when NaN values in the data should be ignored rather than propagating into a NaN result

⚠️

Tip: Pass ddof=1 when your array is a sample meant to estimate a larger population's standard deviation, not the entire population itself — NumPy's default of ddof=0 computes the population version, which several other tools, like pandas, default differently on.

editor.html

import numpy as np

arr = np.array([2, 4, 4, 4, 5, 5, 7, 9])
print(np.std(arr))

localhost:3000

Examples

Example 01Basic Usage

import numpy as np

arr = np.array([2, 4, 4, 4, 5, 5, 7, 9])
print(np.std(arr))

Example 02Advanced Example

import numpy as np

sample = np.array([2, 4, 4, 4, 5, 5, 7, 9])
print(np.std(sample, ddof=0))
print(np.std(sample, ddof=1))

Best Practices

Use ddof=1 for sample-based statistical estimates, matching the convention many statistics courses and other tools default to
Combine std() with mean() to describe a distribution's center and spread together, rather than reporting either alone
Use np.nanstd() when NaN values in the data should be ignored rather than propagating into a NaN result

Interview Question

What does the ddof parameter control in np.std(), and why might you set it to 1?▼

Hint: Think about the difference between describing a whole population versus estimating one from a sample.

ddof stands for 'delta degrees of freedom', and it adjusts the divisor used in the standard deviation calculation from N to N minus ddof. With the default ddof=0, NumPy computes the population standard deviation, dividing by N, which is correct only when your array represents the entire population you care about. Setting ddof=1 instead divides by N-1, giving Bessel's correction, which produces an unbiased estimate of a larger population's standard deviation when your array is actually just a sample drawn from it — the standard convention in most statistical contexts.

Exercises

MediumPractice using np.std() in a real scenario.

View Solution

import numpy as np

arr = np.array([2, 4, 4, 4, 5, 5, 7, 9])
print(np.std(arr))

Frequently Asked Questions

What does the ddof parameter control in np.std(), and why might you set it to 1?▼

Related Functions

np-varnp-meannp-percentile