Having a list of numbers such as [968, 233, 493, 5, 851, 224, 871, 396, 689, 862] I would like to get a rank of each of them, where the highest number gets the lowest rank.
By using scipy.stats.rankdata I get this array([10., 3., 5., 1., 7., 2., 9., 4., 6., 8.]).
The problem is that the ranks are given assuming sorting the list in ascending order. How can I get the ranks assuming descending order of sorting given rankdata doesn't take this as a parameter?
Another way to do it is to multiply each element by (-1):
import numpy as np
import scipy
a = np.array([968, 233, 493, 5, 851, 224, 871, 396, 689, 862])
scipy.stats.rankdata(-a)
It gives:
array([ 1., 8., 6., 10., 4., 9., 2., 7., 5., 3.])
If you do not have to use rankdata you can use numpy argsort.
import numpy as np
inds = np.argsort(arr)
ranks = np.empty_like(inds)
ranks[inds] = np.arange(len(arr),0,-1)
should give you the results you want.
This works by getting the ordering of the elements of the original array using argsort, and then populating a new array using a descending list of numbers in order.
