def run_pa(plan, total_steps=10000000):
partition, chain = set_up_chain(plan, total_steps)
scores = {
key: value
for election in elections
for key, value in get_scores(election).items()
}
# scores['Distance_From_Ideal_Population'] = distance_from_ideal_population
scores['L_minus_1_Polsby-Popper'] = L_minus_1_polsby_popper
initial_scores = {key: score(partition) for key, score in scores.items()}
table = pipe_to_table(chain, scores)
pathlib.Path(f"./plots/{plan}/{now}").mkdir(parents=True, exist_ok=True)
for score in scores:
plt.hist(table[score], bins=50)
plt.title(score.replace('_', ' '))
plt.axvline(x=initial_scores[score], color='r')
plt.savefig(f"./plots/{plan}/{now}/{score}.svg")
plt.close()
metadata = {'plan': plan, 'total_steps': total_steps}
report = {
key: p_value_report(key, table[key], initial_scores[key])
for key in scores if key != 'L_minus_1_Polsby-Popper'
}
return {**metadata, 'p_value_report': report}
def main():
initial_partition = PA_partition()
chain = BasicChain(initial_partition, total_steps=10000)
scores = {
'Mean-Median':
functools.partial(mean_median, proportion_column_name='VoteA%'),
'Mean-Thirdian':
functools.partial(mean_thirdian, proportion_column_name='VoteA%'),
'Efficiency Gap':
functools.partial(efficiency_gap, col1='VoteA', col2='VoteB'),
}
initial_scores = {
key: score(initial_partition)
for key, score in scores.items()
}
table = pipe_to_table(chain, scores)
return {
key: p_value_report(key, table[key], initial_scores[key])
for key in scores
}
def write_p_values(a, b, c, filename=''):
initial_plans = a[0][0]
output = [
outputs.p_value_report(score, a[0][score], initial_plans[score])
for score in a[2]
]
with open(filename, "w") as f:
f.write(json.dumps(output))
def test_p_value_report_gives_the_right_value_when_the_whole_ensemble_is_lower_than_initial(
):
mock_ensemble_scores = [1] * 199 + [17] * 1
initial_plan_score = 17
# We have 199 scores < 17, and 1 score >= 17, so we have 1/200
# higher than the initial score. This should give a p-value of
# sqrt(2 * 1/200) = 1/10.
result = p_value_report('Mock Score', mock_ensemble_scores,
initial_plan_score)
assert result['fraction_as_high'] == 1 / 200
assert result['p_value'] == 1 / 10
assert result['opposite_p_value'] == math.sqrt(2 * 199 / 200)
scores,
display=True,
number_to_display=number_to_display,
bin_interval=bin_interval)
print(steps, " Steps in ", time.time() - start_time, " Seconds")
pscores = dict(scores)
pscores.pop("Node Flipped:")
pscores.pop("Flipped to:")
pscores.pop("All Flips:")
# P-value reports
pv_dict = {
key: p_value_report(key, table[key], initial_scores[key])
for key in pscores
}
# print(pv_dict)
with open(newdir + 'pvals_report_multi.json', 'w') as fp:
json.dump(pv_dict, fp)
print("computed p-values")
# Histogram and trace plotting paths
hist_path = newdir + "chain_histogram_multi_"
trace_path = newdir + "chain_traces_multi_"
# Plots for each election
for i in range(num_elections):
table = pipe_to_table(chain, scores, display=True, number_to_display=number_to_display,
bin_interval=bin_interval)
print(steps," Steps in ", time.time()-start_time," Seconds")
pscores = dict(scores)
pscores.pop("Node Flipped:")
pscores.pop("Flipped to:")
pscores.pop("All Flips:")
# P-value reports
pv_dict = {key: p_value_report(key, table[key], initial_scores[key]) for key in pscores}
# print(pv_dict)
with open(newdir + 'pvals_report_multi.json', 'w') as fp:
json.dump(pv_dict, fp)
print("computed p-values")
# Histogram and trace plotting paths
hist_path = newdir + "chain_histogram_multi_"
trace_path = newdir + "chain_traces_multi_"
# Plots for each election
for i in range(num_elections):
