Last active
December 10, 2015 02:48
-
-
Save josef-pkt/4370540 to your computer and use it in GitHub Desktop.
scoreatpercentiles vectorized
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
# -*- coding: utf-8 -*- | |
"""scoreatpercentiles vectorized | |
Created on Mon Dec 24 09:29:31 2012 | |
Author: Josef Perktold | |
""" | |
import numpy as np | |
def percentiles(x, per, interpolation_method='fraction', issorted=False, | |
axis=0): | |
'''score_at_percentiles vectorized | |
''' | |
TINY = 1e-15 | |
if not issorted: | |
xsorted = np.sort(x, axis=axis) | |
else: | |
xsorted = x | |
scalar = np.isscalar(per) | |
per = np.atleast_1d(per) | |
per_invalid = (per < 0) | (per > 100) | |
if per_invalid.any(): | |
per = np.clip(per, 0, 100) | |
n = (xsorted.shape[axis] - 1) | |
idx = per / 100. * n | |
if interpolation_method == 'lower': | |
idx = np.floor(idx + TINY) | |
elif interpolation_method == 'higher': | |
idx = np.ceil(idx - TINY) | |
elif interpolation_method == 'fraction': | |
pass # keep idx as fraction and interpolate | |
else: | |
raise ValueError("interpolation_method can only be 'fraction', " \ | |
"'lower' or 'higher'") | |
return_shape = list(xsorted.shape) | |
return_shape[axis] = len(per) | |
sumval = np.empty(return_shape) | |
sumval.fill(np.nan) | |
i = idx.astype(int) | |
indexer1 = [slice(None)] * xsorted.ndim | |
indexer1[axis] = np.atleast_1d(i) | |
indexer2 = [slice(None)] * xsorted.ndim | |
indexer2[axis] = np.atleast_1d(np.minimum(i+1, n-1)) | |
weights1 = np.array((i+1 - idx), float) | |
weights2 = np.array((idx - i), float) | |
mask = (i == idx) | |
weights1[mask] = 1 | |
weights2[mask] = 0 | |
wshape = [1] * xsorted.ndim | |
wshape[axis] = len(per) | |
weights1.shape = wshape | |
weights2.shape = wshape | |
result = xsorted[indexer1] * weights1 + xsorted[indexer2] * weights2 | |
result /= (weights1 + weights2) | |
if per_invalid.any(): | |
sl = indexer1 = [slice(None)] * result.ndim | |
sl[axis] = per_invalid | |
result[sl] = np.nan | |
if result.size == 1 and scalar: | |
return result.item() | |
return result | |
x = np.arange(11) | |
np.random.shuffle(x) | |
per = [0, 10, 21, 25, 29.9, 50, 75, 90, 100] | |
interpolation_method = 'fraction' | |
axis = 0 | |
print percentiles(x, per, interpolation_method, axis=axis) | |
print percentiles(x, 50, interpolation_method, axis=axis) | |
print percentiles(x, 10*np.sqrt(3)**2, interpolation_method, axis=axis) | |
print percentiles(np.repeat([x],3,0), per, interpolation_method, axis=1) | |
print percentiles(np.repeat([x],3,0).T, per, interpolation_method, axis=0) | |
print percentiles(np.repeat([x],3,0).T, per, 'lower', axis=0) | |
print percentiles(np.repeat([x],3,0).T, per, 'higher', axis=0) | |
print percentiles(np.repeat([x],3,0).T, [-5, 50, 120], 'higher', axis=0) | |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment