calibva.sampler: Performs Gibbs sampling for calibration

Description Usage Arguments Value

View source: R/calibration.R

Description

calibva.sampler takes in estimation of the underlying cause of death distribution from training data, as well as a transition matrix based on calibration data. Along with the prior values, it will return a list of posterior samples for parameters of interest

Usage

1
2
3
4
calibva.sampler(test.cod, calib.cod = NULL, calib.truth = NULL, causes,
  epsilon, alpha, beta, tau.vec, delta, gamma.init, ndraws, nchains = 1,
  init.seeds = NULL, max.gamma = 75, sample.gamma = TRUE,
  gamma.final = NULL)

Arguments

test.cod

will be a vector of length N, with each entry as the estimated COD (as a character)for indiv. i

calib.cod

is in the same format as test.cod, except for the calibration set

calib.truth

is a character vector with the true COD for each subject in the calibration set

causes

is a character vector with the names of the causes you are interested in. The order of the output vector p will correspond to this vector

epsilon

A numeric value for the epsilon in the prior

alpha

A numeric value for the alpha in the prior

beta

A numeric value for the beta in the prior

tau.vec

A numeric vector for the logsd for the sampling distributions of the gammas

delta

A numeric value for the delta in the prior

gamma.init

A numeric value for the starting value of gammas

ndraws

The number of posterior samples to take within each chain

nchains

The number of chains

init.seeds

An optional numeric vector, of length nchains, with the initial seeds for each chain

max.gamma

The maximum value gamma is allowed to take in posterior samples. Default is 75.

Value

a mcmc.list of length nchains, where each element is a mcmc.object containing the posterior draws for a given chain


jfiksel/CalibratedVA documentation built on July 22, 2019, 7:16 p.m.