negbinom: Bayesian negative binomial regression

In DoktorMike/bayesian-model-zoo: Provides an interface to precompiled useful real life models implemented in Stan

Description

This implements a Bayesian negative binomial regression by using the 2nd parameterization from the Stan manual.

Usage

negbinom(y, x, chains = 2, iter = 1000, allpars = FALSE,
  cores = max(parallel::detectCores() - 1, 1), ...)

Arguments

y	the response variable to uses
x	the matrix of covariates to use
chains	the number of chains to use which defaults to 2
iter	the number of samples to pull which defaults to 1000
allpars	decides if all parameters should be returned or not. FALSE which is the default only gives the mu's, beta's, phi and predicted y_rep
cores	the number of cores to use which defaults to max(parallel::detectCores()-1, 1)
...	other arguments passed to the sampling method in rstan

Value

the stanfit from the sampled model

Examples

## Not run: 
  data<-list(N=nrow(mtcars), y=mtcars$gear, X=mtcars[, c('drat', 'am')])
  sfit<-negbinom(data$y, data$X, iter = 500, chains=2, cores=2, allpars = TRUE)
  library(bayesplot)
  mcmc_combo(as.array(sfit), regex_pars = "tau")
  yrep<-as.matrix(sfit)
  ppc_dens_overlay(data$y, yrep[1:50,grep("y_", colnames(yrep))])

## End(Not run)

DoktorMike/bayesian-model-zoo documentation built on May 23, 2019, 11:32 p.m.