Montecarlo Elections

Class 4

February 12th, 2020

Data Science: Techniques and Applications (DSTA)

A Special lecture

A continuation of the topics seens in the PwD module (probabilities)

Context

opinion polls return a probabilistic measure/and interval
when intervals overlap, different outcomes are possible

(source: El Pais, June 3rd 2018.)

Meta-statistics

Montecarlo simulations

The setup

each result is draw with a probability
Gaussian distribution centered on the meta-statistics
ignore electoral flows (for the moment)

            N = 20000

            # ISSUE: negative values are possible
            votes = rand.normalvariate(mu, sigma)

            print(cand, ' has received ', votes, ' % of votes')

(inefficient) Rejection Sampling to simulate realistic turnouts

print('good election, turnout is ', tot)

winner = max(res, key=lambda k: res[k])

# record victory in candidates 
next(item for item in candidates if item['cand'] == winner)['wins']+=1

[...]

# wins as a percentage
for c in candidates:
    c['wins'] /= N

Results

Who will be president?

Montecarlo Elections Class 4 February 12th, 2020