ccvx_fit: Sample from Concavex model posteriors

In paulmanser/concavex: Concavex Bayesian Dose-Response Modeling

Description

This function performs Gibbs sampling from posterior densities of concavex model parameters as well as transformations of parameters.

Usage

ccvx_fit(ccvx.mod, doses, mu.hat, std.err, n.chains = 4,
  gibbs.samples = 5000, burn.in = 1000, sd.ph3 = NULL,
  n.per.arm.ph3 = NULL)

Arguments

ccvx.mod	JAGS model file as specified by ccvx_build_jags
doses	A vector of dose strengths, with placebo listed first
mu.hat	A vector of parameter estimates for each of the doses
std.err	A vector of standard errors for each of the doses
n.chains	Number of chains used to for Gibbs sampling. Default is 4
gibbs.samples	Number of samples to draw for each chain after burn in. Default is 5000
burn.in	Number of burn in samples to be drawn. Default is 1000
sd.ph3	The standard deviation of the endpoint being assessed in phase 3. Only needs to be specified when ccvx_fit() computes posterior predictive probabilities (i.e. model-based DDCPs)
n.per.arm.ph3	The number of patients per arm being assessed in hypothetical phase 3 study. Only needs to be specified when ccvx_fit() computes posterior predictive probabilities (i.e. model-based DDCPs)

Value

A list with the elements

ccvx.mod	The JAGS code used for Gibbs sampling as generated by ccvx_build_jags or ccvx5_build_jags
jags.samples	output containing Gibbs samples from parameter posteriors generated
coda.samples	output containing coda samples for MCMC diagnostics
doses	A vector of dose strengths, with placebo listed first
mu.hat	A vector of parameter estimates for each of the doses
std.err	A vector of standard errors for each of the doses

Examples

ccvx.mod <- ccvx_build_jags()
ccvx.samples <- ccvx_fit(ccvx.mod, doses = 0:4, mu.hat = c(1, 20, 50, 60, 65), std.err = rep(10, 5))
names(ccvx.samples$jags.samples)
ccvx_plot_fit(ccvx.samples)

paulmanser/concavex documentation built on May 5, 2019, 5:53 p.m.