calculations for hypothetical scenarios in auditing the 2106 Australian senatorial election, Tasmanian portion

aus_16_audit.py

# Calculate various hypotheticals for the 2016 Australian Senatorial election, for Tasmanian results.
# P.B. Stark, 17 September 2016

from __future__ import division, print_function
import math
import numpy as np
import scipy as sp
from scipy import stats  # distributions
from scipy import special # special functions
from scipy import random # random variables, distributions, etc.
from scipy.optimize import brentq
from scipy.stats import (binom, hypergeom)
import permute.utils

cl = 0.95  # confidence level for these calculations
alpha = 1-cl # corresponding risk limit

valid_votes = 339159   # total votes in Tasmania
invalid_votes = 12221  # invalid votes, blank ballots, etc.
ballots = valid_votes + invalid_votes # total ballots

marginv = 141 # votes separating runner-up from getting a seat
margindv = marginv/ballots  # "diluted" margin in votes 
                            # (i.e., if an error can decrease the margin by 1 or 2)

marginb = 71  # minimum number ballots that would need to have errors to alter the outcome
margindb = marginb/ballots  # "diluted" margin in ballots

print('upper bound on the diluted margin in ballots:', margindb)
print('upper bound on the diluted margin in votes:', margindv)

# Measured risk based on observing no errors.
# for sample of size n, chance of observing no errors if the true error rate is large
# enough to change the outcome is (1-marginp)^n

smallest_sample = 1000
largest_sample = 16000
sample_increment = 1000

print('sample size, measured risk if no errors are observed in the sample:')
for n in np.arange(smallest_sample, largest_sample+1, sample_increment):
    print(n, (1-margindb)**n)

# upper bounds on the error rate for various sample sizes
x = 0

print('sample size, upper 95% confidence bound for error rate if no errors are observed in the sample:')
for n in np.arange(smallest_sample, largest_sample+1, sample_increment):
    print(n, permute.utils.binom_conf_interval(n, x, cl=cl, alternative="upper"))
    
# Suppose that x errors are observed in a sample of 2500 ballots.
# lower confidence bounds on population error rate for various k

print('errors observed in a sample of 2500 ballots, lower 95% confidence bound on error rate:')
n = 2500
for x in np.arange(1, 11):
    print(x, permute.utils.binom_conf_interval(n, x, cl=cl, alternative="lower")) 
    
# Initial sample size for a risk-limiting audit, assuming the true error rates are zero

def minSampleSize(ballots, u, alpha=0.05, gamma=0.95):
    '''
    find smallest sample size for risk-limit alpha, using cushion gamma \in (0,1)
    1/alpha = (gamma/(1-1/(ballots*u))+1-gamma)**n
    Input: 
        ballots: number of ballots cast in the contest
        u:       upper bound on overstatement per ballot
        gamma:   hedge against finding a ballot that attains the upper bound. Larger values give
                 less protection
        alpha:   risk limit
    '''
    return math.ceil(math.log(1.0/alpha) / math.log(gamma/(1.0-1.0/(ballots*u)) + 1.0 - gamma))
    
print('Initial sample size for RLA with risk limit', alpha, ':', \
      minSampleSize(ballots, 2/marginv, alpha=alpha))

Thanks for this code!
Running it in I get ImportError: No module named permute.utils. The Python code for the permute module seems to be described in http://statlab.github.io/permute/permute.pdf
Is the code available somewhere? Or at least the permute.utils.binom_conf_interval() method?
Also, is it written for Python 3, Python 2, or both?