ELECTIONS POLLS
Silas Liu - Sep. 12, 2021
R, Bayesian Modeling
Predicting the outcome of an election is one of many polls applications, maybe the most common one. However the analysis of its data requires proper modeling and probabilistic analysis.
Bayesian statistics allow us to assume a certain probability, as the most favored in an election, as a random variable and not a fixed value. This allows us to use hierarchical models to describe variability at different levels, in order to estimate the probability.
The case study shown here analyzes the 2016 U.S.presidential elections, with real poll data. It illustrates how the polls of popular vote and electoral college missed the predictions, of Clinton winning over Trump and shows the numbers behind it.
The next step of this analysis is located in the next section, where we apply machine learning regression to fit the data.