jwhandley / Bayesian-Poll-of-Polls

Stan code for creating a Bayesian poll of polls for multiparty systems

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Bayesian-Poll-of-Polls

Stan code for creating a Bayesian poll of polls for multiparty systems

How to use

The stan model expects the following data:

  • T: an integer representing the total number of time steps in the model
  • J: an integer representing the total number of parties in the model
  • N: integer number of observations in the data (not necessarily the same as T)
  • P: integer number of unique pollsters
  • t: integer time step for each observation. Assumed to go from 1 to T.
  • vi: N by J matrix of poll results. Set to -1 if a party is missing.
  • pollster: integer id for the pollster of each observation. Should range from 1 to P
  • size: vector of sample sizes for each poll
  • vote0: vector used to initialize the state variable; should contain one value per party and sum to 1

Once you have the data in this format, you can compile and run the code with your preferred Stan interface. I have a couple of examples that use cmdstanr to produce graphs from Europe Elects polling data in the scripts folder.

About

Stan code for creating a Bayesian poll of polls for multiparty systems

License:MIT License


Languages

Language:R 69.3%Language:Stan 30.7%