Python:NumPy .quantile()

The .quantile() function in NumPy returns the qth quantile of an array along a specified axis. Quantiles are the division points that separate a data set into equal probabilities. For example, the 25th quantile is the point which 25% of the data set falls below.

Syntax

numpy.quantile(a, q, axis=None, out=None, overwrite_input=False, method='linear', keepdims=False, weights=None)

Parameters:

• a: The input array containing the data to compute the quantiles from.
• q: The quantile(s) to compute. This can be a float or array-like of floats between 0 and 1, where 0.5 represents the median.
• axis (Optional): The axis or axes on which to calculate the quantile. axis=0 computes along columns, and axis=1 computes along rows. If set to None (default), the input is flattened before computation.
• out (Optional): Specifies a different array in which to place the result. It must have the same shape as the expected result.
• overwrite_input (Optional): If True, the input array a may be modified to save memory. Default is False.
• method (Optional): The method used to calculate the quantile. The default is 'linear'. Valid options include: 'inverted_cdf', 'averaged_inverted_cdf', 'closest_observation', 'interpolated_inverted_cdf', 'hazen', 'weibull', 'median_unbiased', and 'normal_unbiased'. keepdims (Optional): If True, the reduced axes are retained with size one, maintaining the number of dimensions in the output. weights (Optional): An array of weights corresponding to values in a, used to influence the quantile calculation. This parameter is only supported by the 'inverted_cdf' method. The shape of weights must either match a, or be 1-dimensional with a length equal to a when flattened.

Return value:

The .quantile() function returns the qth quantile(s) of an array as a NumPy array (ndarray) or a scalar (float64) if the result is a single value.

Example: Computing multiple quantiles from data

The following example creates an array and then uses .quantile() to calculate various quantiles from the data:

import numpy as np

a = np.array([[0,1,2],[3,4,5]])

print(np.quantile(a, .25))
# Computes the 25th quantile along a flattened axis

print(np.quantile(a, .5, axis=0))
# Computes the 50th quantile along the vertical axis

print(np.quantile(a, .5, axis=1))
# Computes the 50th quantile along the horizontal axis

print(np.quantile(a, .75, axis=1, keepdims=True))
# Computes the 75th quantile along the horizontal axis, while retaining the original dimensions of the input array

Copy to clipboard
Copy to clipboard

This code produces the following output:

1.25
[1.5 2.5 3.5]
[1. 4.]
[[1.5]
 [4.5]]

Copy to clipboard
Copy to clipboard

Codebyte Example

The following codebyte example computes various quantiles for an input array, a:

Visit us
Visit us
Hide code
Code
Output
Hide output
Loading...
Copy to your clipboard
Copy Codebyte
Run