np.percentile returns different median from np.median when inf is present

Question

When inf values are present in an array, under certain conditions np.percentile can return NaN as the median, whereas np.median can return a finite value.

>>> import numpy as np
>>> np.percentile([np.inf, 5, 4], [10, 20, 30, 40, 50, 60, 70, 80, 90])
/Users/tom/miniconda3/envs/alldev/lib/python3.7/site-packages/numpy-1.16.0.dev0+45718fd-py3.7-macosx-10.7-x86_64.egg/numpy/lib/function_base.py:3947: RuntimeWarning: invalid value encountered in multiply
  x2 = take(ap, indices_above, axis=axis) * weights_above
array([4.2, 4.4, 4.6, 4.8, nan, inf, inf, inf, inf])
>>> np.median([np.inf, 5, 4])
5.0

In this case, np.median is able to correctly return 5.0 as the median value, whereas np.percentile returns NaN for the 50th percentile.

That's because the functionality of those functions are not the same, not even close. You should check the source codes. — Mazdak, Dec 27 '18 at 15:00
No, It's a good question. You can provide the answer after by yourself whenever you found it or my be someone else does that too. — Mazdak, Dec 27 '18 at 15:07

score 0 · Accepted Answer · answered Mar 29 '21 at 19:18

the first parameter is the data, the second parameter is the confidence intervals. you not allowed to put an non number in the confidence interval

data=[10, 20, 30, 40, 50, 60, 70, 80, 90]
confidence=[5, 4]
results=np.percentile(data,confidence )

print(results)

output

 array([14. , 13.2])

values 13.2 through 14 will yield a 4 to 5 percent confidence interval

np.percentile returns different median from np.median when inf is present

1 Answers1