DAC (version 0.1.1)

DAC.normal: DAC For Normal Benchmark Prior And Expert Priors

Description

Calculates DAC values for multiple experts that provide prections in the form of a prior. The benchmark is set to be a normal distribution with user specified boundaries.

Usage

DAC.normal(from, to, by, data, priors, mean.bench, sd.bench, n.iter = 10000)

Arguments

from

Lower bound of the parameter space that is to be evaluated, as in the seq function of the base package.

to

Upper bound of the parameter space that is to be evaluated, as in the seq function of the base package.

by

Step size by which the defined parameter space is mapped out, as in the seq function of the base package.

data

A vector of your data points.

priors

A matrix of densities with in each column a density of a specific prior mapped on the paramater space that is equal to the parameter space that is supplied using the from, to, by statements. E.g. the parameter space runs from -10 to 10 in steps of 0.01 than your density of a standard normal distribution should be obtained using dnorm(x = seq(from = -10, to = 10, by = 0.01), mean = 0, sd = 1). The first column will thus describe this density using 2001 rows and all other columns should use the same density mapping to the parameter space.

mean.bench

Mean of the benchmark prior.

sd.bench

SD of the benchmark prior.

n.iter

The number of iterations that is used to obtain the posterior distribution of the data and the benchmark prior note that only half of these iterations will be used to obtain samples, the other half is used for adaptation and burnin.

References

Bousquet, N. (2008) Diagnostics of prior-data agreement in applied bayesian analysis. Journal of Applied Statistics, 35(9), 1011-1029.

Veen, D., Stoel, D., Schalken, N., van de Schoot, R. (2017). Using the Data Agreement Criterion to Rank Experts' Beliefs. arXiv preprint arXiv:1709.03736.

Examples

# NOT RUN {
y <- Turnover$actual.data
from <- -500 
to <- 500
by <- .01
xx <- seq(from, to, by)
priors <- matrix(NA,ncol=4,nrow=length(xx))
for(i in 1:4){
  priors[,i] <- dnorm(xx, Turnover$expert.priors[i ,1], Turnover$expert.priors[i, 2])
}
# }
# NOT RUN {
out <- DAC.normal(from = from, to = to, by = by, data = Turnover$actual.data, 
                   priors = priors ,mean.bench = 0, sd.bench = 100, n.iter = 1000)
# }
# NOT RUN {
# }