CODEWITHSUNDEEP
pythonnumpy
Hjin
Hjin

Reputation: 322

How to get nth max correlation coefficient and its index by using numpy?

I compute correlation coefficient like this (its just example):

a = np.array([[1, 2, 3],
     [4, 7, 9],
     [8, 7, 5]])

corr = np.corrcoef(a)

The result is a correlation matrix.

The question is how to get 1st, 2nd (or nth) largest coefficient?

And its index? like [0,1] and [2,1]

Upvotes: 1

Views: 1143

Answers (1)

rednafi
rednafi

Reputation: 1731

Let's say you have a NumPy array and you computed the correlation coefficient like this:

import numpy as np

a = np.array([[1, 2, 3],
     [4, 7, 9],
     [8, 7, 5]])

corr = np.corrcoef(a)

Now flatten the array, take the unique coefficients and sort the flattened array:

flat=corr.flatten()
flat = np.unique(flat)

The flat array looks like this:

>> array([-0.98198051, -0.95382097,  0.99339927,  1. ])

Now to pick nth largest element, just pick the right index:

largest = flat[-1]
second_largest = flat[-2]

print(largest)
print(second_largest)

>> 1.0
>> 0.9933992677987828

To find the indices of the corresponding coefficient:

result = np.where(corr == largest)
indices = np.array(result)
print(indices)

This prints out the following array. So the indices where the largest coefficient occurs are (0,0), (1,1) and (2,2).

>> array([[0, 1, 2],
       [0, 1, 2]])

Upvotes: 3

Related Questions