{{ message }}

Instantly share code, notes, and snippets.

# tomron/sprt.py

Last active Apr 30, 2019
Sequential probability ratio test implementation (https://en.wikipedia.org/wiki/Sequential_probability_ratio_test) for exponential distribution. Usage - `t = sprt.SPRT(0.05, 0.8, 1, 2); t.test([1, 2, 3, 4, 5])`
 import numpy as np """ Implements Sequential probability ratio test https://en.wikipedia.org/wiki/Sequential_probability_ratio_test """ class SPRT: def __init__(self, alpha, beta, mu0, mu1): """ Initilize test parameters """ self.alpha = alpha self.beta = beta self.mu0 = mu0 self.mu1 = mu1 def get_bounds(self): """ Get rejection bounds """ bound1 = np.log(self.beta/(1-self.alpha)) bound2 = np.log((1-self.beta)/self.alpha) return (bound1, bound2) def test(self, arr): """ Test hypothesis on a given array Output: (-1, idx) if H0 was accepted and in which index, (1, idx) if H1 was accepted and in which index, (0, -1) continue monitoring """ ratio = np.log(self.mu1/self.mu0) coefficient = (self.mu1 - self.mu0)/(self.mu1 * self.mu0) bound1, bound2 = self.get_bounds() prev_s = -ratio + coefficient * arr[0] for i in range(1, len(arr)+1): curr_s = prev_s - ratio + coefficient * arr[i] if (curr_s <= bound1): return (-1, i) elif (curr_s => bound2): return (1, i) prev_s = curr_s return (0, -1)
to join this conversation on GitHub. Already have an account? Sign in to comment