검정

import numpy as np
from empiricaldist import Pmf
from scipy.stats import binom

k, n = 140, 250
like_fair = binom.pmf(k, n, 0.5)

ramp_up = np.arange(50)
ramp_down = np.arange(50, -1, -1)
a = np.append(ramp_up, ramp_down)

xs = np.linspace(0, 1, 101)
triangle = Pmf(a, xs)
triangle.normalize()
biased_triangle = triangle.copy()
biased_triangle[0.5] = 0
biased_triangle.normalize()

likelihood = binom.pmf(k, n, xs)
like_triangle = np.sum(biased_triangle * likelihood)

k = like_fair / like_triangle
k, k / (k + 1)

(1.1970766535647133, 0.5448497446015277)

• 250번 중 140번이 앞면이 나왔는데도 여전히 공정할 확률이 높다. (왜?)