hb_estimation: Bayesian hierarchical model to estimate proportion of votes...

In cotecora-team-2/quickcountmx: Quick-count estimation package for the 2021 Mexican election

Bayesian hierarchical model to estimate proportion of votes allocated to each party

Description

Compute point estimate and credible intervals for each candidate.

Usage

hb_estimation(
  data_tbl,
  stratum,
  id_station,
  sampling_frame,
  parties,
  covariates,
  prop_obs = 0.995,
  seed = NULL,
  return_fit = FALSE,
  num_iter = 200,
  num_warmup = 200,
  adapt_delta = 0.8,
  max_treedepth = 10,
  chains = 3,
  sig_figs = 6,
  model = "mlogit-corr",
  nominal_max = 1000,
  threads_per_chain = 1,
  inv_metric = NULL,
  erase_output_files = TRUE
)

Arguments

data_tbl	tibble of sample data.
stratum	Unquoted variable indicating the stratum for each polling station.
id_station	Unquoted variable indicating the id for each polling station.
sampling_frame	tibble of sampling frame with stratum variable (named exactly as in data_tbl) and covariates.
parties	Unquoted variables indicating the number of votes in each polling station for each candidate.
covariates	Unquoted variables indicating the covariates in each polling station.
prop_obs	Proportion of size of observed sample to total designed sample.
seed	integer value used to set the state of the random number generator (optional).
return_fit	Returns summary if FALSE (default), otherwise return cmdstanr fit
num_iter	Number of post warmup iterations
num_warmup	Number of warmup iterations
adapt_delta	The adaptation target acceptance statistic (default 0.80.
max_treedepth	The maximum allowed tree depth for the NUTS engine (default 10)
chains	Number of chains (will be run in parallel)
sig_figs	Number of significant figures for Stan output
model	One of "mlogit" (the default) or "logit"
nominal_max	Maximum number of nominal count for stations. Used for stations without fixed nominal list.
threads_per_chain	Number of threads per chain to split calculation of log-posterior
inv_metric	vector of inverse metric diagonal for the model. Default is NULL
erase_output_files	If TRUE, output files are erased after fitting

Details

Posterior simulations of parameters are computed using stan, and each party's votes are simulated for every polling station (logit model) or with a softmax link for the default (mlogit model). There is one independent model for each party, and proportions are calculated from posterior simulations of total votes.

Value

A list with model fit (if return_fit=TRUE), a tibble estimates including point estimates for each party (median) and limits of credible intervals, and a vector inv_metric for the model

cotecora-team-2/quickcountmx documentation built on July 17, 2025, 5:14 a.m.