Project implemented in Python to simulate and make "prediction" about the 2016 presidential election.(well, 3 months after the election). The idea is very simple: assuming that the outcome of the presidential race at each state follows a Bernalloui distribution where Donald Trump (now president-elect) has the probability of winning p , we can iterate the race in each state over a large amount of time and compute the mean electoral votes.
The question is: how do we find p?
Two methods are used:
- Regression based on Washington Post pre-election poll (adjusted by the 2012 biased)
- Bayesian inference with Metropolis-Hasting algorithm
These instructions will get you a copy of the project up and running on your local machine for development and testing purposes. See deployment for notes on how to deploy the project on a live system.
Project uses Python and many relevant statistical, graphical libaries, notably:
- panda
- numpy
- matplotlib
Information on how to install these packages can be found only. I highly recommend using Anaconda platform.
conda install pandas
R is used for Bayesian simulation, mainly due to its computational prowess. I will soon update a code in Python.
I prefer using Jupyter notebook as it retains the logical flow of the code and the way I broke the projects down into smaller problems.
Very simple use, just run notebook.ipynb cell by cell. At some point, you will have to start using R Bayesian_MCMC.Rto conduct Bayesian MCMC.
I will soon add a detailed discussion on the results of the projects.
The project allows you to run a simulation of 10,000 election races and present data in a nice histogram and a choropleth map
- A nice histogram that presents results from 10,000 simulation:
Wtih a less than 3% chance of winning, who would think that he is now the President-elect?
- A chloropeth math that presents probability of Trump winning in each state:
- Tuan Nguyen Doan - Initial work - tuangauss
This is a self-learning project and I hope to learn from the expertise of the community. Please reach out to me if you have any suggestion or ideas.
This is a self-learning project and I am proud to present the following sources as my reference (and inspiration):
- Harvard Open Course CS109: Data Science (the idea of this project comes from one of the problem sets)
- Open-resourced code: for choropleth math, basic question, run_time issues
- Yale Stat 238
- Five-thirty eight
- Open poll data from here