predict_pg_stlm_overdispersed: Bayesian Polya-gamma regression prediction
In jtipton25/pgR: Bayesian Polya-Gamma Regression

View source: R/predict-pg_stlm_overdispersed.R

predict_pg_stlm_overdispersed

R Documentation

Bayesian Polya-gamma regression prediction

Description

this function generates predictions from the Bayesian multinomial regression using Polya-gamma data augmentation

Usage

predict_pg_stlm_overdispersed(
  out,
  X,
  X_pred,
  locs,
  locs_pred,
  corr_fun,
  shared_covariance_params,
  progress = TRUE,
  verbose = FALSE,
  posterior_mean_only = TRUE
)

Arguments

`out`	is a list of MCMC outputs from pg_stlm_overdispersed()
`X`	is a n x p matrix of covariates at the observed locations.
`X_pred`	is a n_{pred} x p matrix of covariates at the locations where predictions are to be made.
`locs`	is a n x 2 matrix of locations where observations were taken.
`locs_pred`	is a n_pred x 2 matrix of locations where predictions are to be made.
`corr_fun`	is a character that denotes the correlation function form. Current options include "matern" and "exponential".
`shared_covariance_params`	is a logicial input that determines whether to fit the spatial process with component specifice parameters. If TRUE, each component has conditionally independent Gaussian process parameters theta and tau2. If FALSE, all components share the same Gaussian process parameters theta and tau2.
`progress`	is a logicial input that determines whether to print a progress bar.
`verbose`	is a logicial input that determines whether to print more detailed messages.
`posterior_mean_only`	is a logical input that flags whether to generate the full posterior predictive distribution (`posterior_mean_only = FALSE`) or just the posterior predictive distribution of the mean response (`posterior_mean_only = TRUE`). For large dataset, the full posterior predictive distribution can be expensive to compute and the posterior distribution of the mean response is much faster to calculte.