np.corrcoef() computes the Pearson correlation coefficient matrix between two or more variables, measuring the strength and direction of their linear relationship.

Syntax

np.corrcoef(x, y=None)

Deep Dive Course

The Pearson correlation coefficient ranges from -1, a perfect negative linear relationship, to 1, a perfect positive linear relationship, with 0 indicating no linear relationship at all. np.corrcoef(x, y) returns a 2x2 matrix, since correlating two variables produces a correlation of each with itself, always exactly 1, on the diagonal, and with the other, the off-diagonal values, which are identical to each other, since correlation is symmetric. It specifically measures linear relationships — two variables can be strongly related in a non-linear way, like a perfect parabola, while still showing a correlation coefficient near 0.

1Understanding np.corrcoef()

💡

A correlation coefficient near 0 doesn't mean two variables are unrelated — it only means they have no strong linear relationship; a perfectly curved, non-linear relationship can still show a correlation near 0 despite being a very predictable relationship.

editor.html

import numpy as np

x = np.array([1, 2, 3, 4, 5])
y = np.array([2, 4, 6, 8, 10])
print(np.corrcoef(x, y))

localhost:3000

2Practical Example

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

editor.html

import numpy as np

hours_studied = np.array([1, 2, 3, 4, 5])
test_scores = np.array([50, 55, 65, 70, 90])
correlation = np.corrcoef(hours_studied, test_scores)[0, 1]
print(round(correlation, 3))

localhost:3000

3Best Practices

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

1. Remember correlation measures only linear relationships — plot the data or check for non-linear patterns before concluding two variables are unrelated based on a low correlation coefficient alone

2. Extract the specific off-diagonal value when you only need the correlation between two variables, rather than reading the whole matrix

3. Never assume correlation implies causation — a high correlation coefficient only describes a statistical relationship, not that one variable causes changes in the other

⚠️

Tip: A correlation coefficient near 0 doesn't mean two variables are unrelated — it only means they have no strong linear relationship; a perfectly curved, non-linear relationship can still show a correlation near 0 despite being a very predictable relationship.

editor.html

import numpy as np

x = np.array([1, 2, 3, 4, 5])
y = np.array([2, 4, 6, 8, 10])
print(np.corrcoef(x, y))

localhost:3000

Examples

Example 01Basic Usage

import numpy as np

x = np.array([1, 2, 3, 4, 5])
y = np.array([2, 4, 6, 8, 10])
print(np.corrcoef(x, y))

Example 02Advanced Example

import numpy as np

hours_studied = np.array([1, 2, 3, 4, 5])
test_scores = np.array([50, 55, 65, 70, 90])
correlation = np.corrcoef(hours_studied, test_scores)[0, 1]
print(round(correlation, 3))

Best Practices

Remember correlation measures only linear relationships — plot the data or check for non-linear patterns before concluding two variables are unrelated based on a low correlation coefficient alone
Extract the specific off-diagonal value when you only need the correlation between two variables, rather than reading the whole matrix
Never assume correlation implies causation — a high correlation coefficient only describes a statistical relationship, not that one variable causes changes in the other

Interview Question

Why can two variables with a strong, perfectly predictable non-linear relationship still show a correlation coefficient close to 0?▼

Hint: Think about what specifically the Pearson correlation coefficient is designed to measure.

The Pearson correlation coefficient specifically quantifies the strength of a linear relationship — how well a straight line fits the relationship between the two variables. A relationship like y equals x squared, evaluated symmetrically around zero, is completely predictable and deterministic, but as x increases from negative to positive, y first decreases then increases, so there's no consistent straight-line trend across the whole range, and the linear correlation coefficient ends up near 0 despite the strong, exact non-linear relationship. This is exactly why correlation should never be the only tool used to check whether two variables are related — visualizing the data, or using a measure designed for non-linear relationships, is necessary to catch this kind of pattern.

Exercises

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

View Solution

import numpy as np

x = np.array([1, 2, 3, 4, 5])
y = np.array([2, 4, 6, 8, 10])
print(np.corrcoef(x, y))

Frequently Asked Questions

Why can two variables with a strong, perfectly predictable non-linear relationship still show a correlation coefficient close to 0?▼

Related Functions

np-meannp-stdnp-var