BCEs0 (version 1.1-1)

bces0: Bayesian Cost-Effectiveness models in the presence of structural zeros

Description

Writes a model file encoding the distributional assumptions, calls JAGS in background and perform the Bayesian analysis of the selected model.

Usage

bces0(data,dist.c=c("gamma","logn","norm"),
	dist.e=c("beta","gamma","bern","norm"),
	w=1e-6,W=1e-6,n.iter=10000,n.burnin=5000,
	n.chains=2,robust=TRUE,model.file="model.txt")
"bces0"(data,dist.c=c("gamma","logn","norm"),
	dist.e=c("beta","gamma","bern","norm"),
	w=1e-6,W=1e-6,n.iter=10000,n.burnin=5000,
	n.chains=2,robust=TRUE,model.file="model.txt")

Arguments

data

A named list including values for the variables e0 (measure of effectiveness for the subjects in treatment arm t=0), e1 (effectiveness for the subjects in t=1), c0 (individual costs in t=0), c1 (individual costs in t=1), H.psi and H.zeta (vectors of fixed hyperparameters for the prior in the positive cost groups. If only one value is passed as argument, then BCEs0 assumes that this is to be used for both treatments being considered). Additional optional elements are X0 (a matrix of covariates for t=0) and X1 (a matrix of covariates for t=1) that can be used to estimate the selection model for null costs

dist.c

A text string defining the selected distribution for the costs. Available options are Gamma ("gamma"), log-Normal ("logn") and Normal ("norm")

dist.e

A text string defining the selected distribution for the measure of effectiveness. Available options are Beta ("beta"), Gamma ("gamma"), Bernoulli ("bern") and Normal ("norm")

w

A parameter used to characterise the mean of the degenerate distribution for the structural zeros (default = 0.000001)

W

A parameter used to characterise the standard deviaiton of the degenerate distribution for the structural zeros (default = 0.000001)

n.iter

Number of iterations to be run in JAGS (default = 10000)

n.burnin

Number of iterations to be used as burn-in for the MCMC procedure (default = 5000)

n.chains

Number of Markov chains to be run (default = 2)

robust

A string indicating whether a robust model should be chosen for the patter model. If TRUE (default), then the regression coefficients are modelled using a Cauchy(0,2.5) distribution. If FALSE, then a vague Normal prior is used

model.file

A string with the name of the txt file to which the JAGS code is saved. Default is model.txt.

Value

An object containing the following elements

mod

A "rjags" objects with the results of the MCMC simulations run using JAGS

params

A vector including the parameters being monitored

dataJags

A list contaning the data needed to run the MCMC simulations

inits

A function used to initialise the random nodes in the model

References

Baio G. (2013). Bayesian models for cost-effectiveness analysis in the presence of structural zero costs. http://arxiv.org/pdf/1307.5243v1.pdf

Examples

data(acupuncture)
m <- bces0(data,dist.c="gamma",dist.e="beta",n.iter=1000,n.burnin=500,n.chains=2)
print(m)
plot(m)