Probability test

fastai

Published

October 7, 2020

Probability test

Calculating probabilities of values in relation to threshold values.

The code is based upon the respective example from Data Science from Scratch.

Libraries

from typing import Tuple
import math as m

import pandas as pd
import numpy as np

Probability below a threshold

$ p = {2} $

where

$ z = {} _0^z e^{-t2} dt $

is the error function.

def calc_normal_cdf(x: float, mu: float = 0, sigma: float = 1) -> float:
    return (1 + m.erf((x - mu) / m.sqrt(2) / sigma)) / 2

normal_probability_below = calc_normal_cdf

for i in [n / 10 for n in range(-10, 10 + 1, 1)]:
    print("\t".join([str(i), f"{normal_probability_below(i):.4f}"]))

-1.0    0.1587
-0.9    0.1841
-0.8    0.2119
-0.7    0.2420
-0.6    0.2743
-0.5    0.3085
-0.4    0.3446
-0.3    0.3821
-0.2    0.4207
-0.1    0.4602
0.0 0.5000
0.1 0.5398
0.2 0.5793
0.3 0.6179
0.4 0.6554
0.5 0.6915
0.6 0.7257
0.7 0.7580
0.8 0.7881
0.9 0.8159
1.0 0.8413

Probability above a threshold

def normal_probability_above(lo: float, mu: float = 0, sigma: float = 1) -> float:
    return 1 - normal_probability_below(lo, mu, sigma)

for i in [n / 10 for n in range(-10, 10 + 1, 1)]:
    print("\t".join([str(i), f"{normal_probability_above(i):.4f}"]))

-1.0    0.8413
-0.9    0.8159
-0.8    0.7881
-0.7    0.7580
-0.6    0.7257
-0.5    0.6915
-0.4    0.6554
-0.3    0.6179
-0.2    0.5793
-0.1    0.5398
0.0 0.5000
0.1 0.4602
0.2 0.4207
0.3 0.3821
0.4 0.3446
0.5 0.3085
0.6 0.2743
0.7 0.2420
0.8 0.2119
0.9 0.1841
1.0 0.1587

Probability between thresholds

domain = np.arange(-10, 10 + 1, 2) / 10
domain

array([-1. , -0.8, -0.6, -0.4, -0.2,  0. ,  0.2,  0.4,  0.6,  0.8,  1. ])

def normal_probability_between(lo: float, hi: float, mu: float = 0, sigma: float = 1) -> float:
    return normal_probability_below(hi, mu, sigma) - normal_probability_below(lo, mu, sigma)

probabilities_between = pd.DataFrame()

for i in domain:
    for j in domain:
        probabilities_between.loc[i, j] = normal_probability_between(i, j)

probabilities_between = pd.DataFrame(np.triu(probabilities_between), index=domain, 
columns=domain).replace(0, np.NaN)

for i in domain:
    probabilities_between.loc[i, i] = 0

probabilities_between

	-1.0	-0.8	-0.6	-0.4	-0.2	0.0	0.2	0.4	0.6	0.8	1.0
-1.0	0.0	0.0532	0.115598	0.185923	0.262085	0.341345	0.420604	0.496766	0.567092	0.629489	0.682689
-0.8	NaN	0.0000	0.062398	0.132723	0.208885	0.288145	0.367404	0.443566	0.513891	0.576289	0.629489
-0.6	NaN	NaN	0.000000	0.070325	0.146487	0.225747	0.305007	0.381169	0.451494	0.513891	0.567092
-0.4	NaN	NaN	NaN	0.000000	0.076162	0.155422	0.234681	0.310843	0.381169	0.443566	0.496766
-0.2	NaN	NaN	NaN	NaN	0.000000	0.079260	0.158519	0.234681	0.305007	0.367404	0.420604
0.0	NaN	NaN	NaN	NaN	NaN	0.000000	0.079260	0.155422	0.225747	0.288145	0.341345
0.2	NaN	NaN	NaN	NaN	NaN	NaN	0.000000	0.076162	0.146487	0.208885	0.262085
0.4	NaN	NaN	NaN	NaN	NaN	NaN	NaN	0.000000	0.070325	0.132723	0.185923
0.6	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	0.000000	0.062398	0.115598
0.8	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	0.000000	0.053200
1.0	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	0.000000

Probabilities outside a threshold range

def normal_probability_outside(lo: float, hi: float, mu: float = 0, sigma: float = 1) -> float:
    return 1 - normal_probability_between(lo, hi, mu, sigma)

probabilities_outside = pd.DataFrame()

for i in domain:
    for j in domain:
        probabilities_outside.loc[i, j] = normal_probability_outside(i, j)

pd.DataFrame(np.triu(probabilities_outside), index=domain, columns=domain).replace(0, np.NaN)

	-1.0	-0.8	-0.6	-0.4	-0.2	0.0	0.2	0.4	0.6	0.8	1.0
-1.0	1.0	0.9468	0.884402	0.814077	0.737915	0.658655	0.579396	0.503234	0.432908	0.370511	0.317311
-0.8	NaN	1.0000	0.937602	0.867277	0.791115	0.711855	0.632596	0.556434	0.486109	0.423711	0.370511
-0.6	NaN	NaN	1.000000	0.929675	0.853513	0.774253	0.694993	0.618831	0.548506	0.486109	0.432908
-0.4	NaN	NaN	NaN	1.000000	0.923838	0.844578	0.765319	0.689157	0.618831	0.556434	0.503234
-0.2	NaN	NaN	NaN	NaN	1.000000	0.920740	0.841481	0.765319	0.694993	0.632596	0.579396
0.0	NaN	NaN	NaN	NaN	NaN	1.000000	0.920740	0.844578	0.774253	0.711855	0.658655
0.2	NaN	NaN	NaN	NaN	NaN	NaN	1.000000	0.923838	0.853513	0.791115	0.737915
0.4	NaN	NaN	NaN	NaN	NaN	NaN	NaN	1.000000	0.929675	0.867277	0.814077
0.6	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	1.000000	0.937602	0.884402
0.8	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	1.000000	0.946800
1.0	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	1.000000