Prior Probability Functions of the Standard and Truncated Distribution.

ggdmcPrior

This package provides tools for specifying and evaluating standard and truncated probability distributions, with support for log-space computation and joint distribution specification. It is designed to integrate seamlessl with the 'ggdmc' package, enabling Bayesian computation for cognition models. The package includes utilities for density calculation, sampling, and visualisation, making it easier to define priors and assess model behaviour in hierarchical Bayesian framework.

Getting Started

Although the package can be used independently, it is primarily intended to support the 'ggdmc' package, and is best used alongside other packages in the 'ggdmc' ecosystem.

cat("\n--------- Printing the joint prior distribution in C++-------\n")
p0 <- c(A = 0.15, B = 0.45, mean_v = 2.25, sd_v = 0.15, t0 = 0.2)
p1 <- rep(0.1, 5)
names(p1) <- names(p0)


p_prior <- BuildPrior(
    p0 = p0,
    p1 = p1,
    lower = rep(NA, 5),
    upper = rep(NA, 5),
    log_p = rep(TRUE, 5),
    dist = c("tnorm", "tnorm", "tnorm", "norm", "tnorm")
)


prior_S4 <- set_priors(p_prior = p_prior)
names(p_prior)

# Test print -----------------
res <- print_prior(p_prior)

# Test dprior -----------------
parameters <- seq(0.1, 0.5, 0.1)
res0 <- dprior(p_prior, parameters)

set.seed(123)
result1 <- rprior(p_prior, 1)
set.seed(123)
result2 <- rprior(p_prior, 2)

# Use the beta distribution to create uniform densities
# lower and upper set the bounds. If lower is NA, it will be set to 0.
# If upper is NA, it will be set to 1.
p_prior <- BuildPrior(
    p0 = c(A = 1, B = 1, mean_v = 1, sd_v = 1, t0 = 1),
    p1 = rep(1, 5),
    lower = rep(0, 5),
    upper = rep(5, 5),
    dist = rep("beta", 5),
    log_p = rep(TRUE, 5)
)
prior_S4 <- set_priors(p_prior = p_prior)

p0 <- plot_prior(p_prior)

The 'ggdmcPrior' package supports six standard statistical distributions—normal, truncated or regular normal, gamma, lognormal, Cauchy, and uniform—making it easy to define and evaluate priors for Bayesian cognitive modeling.

p0 <- c(A = 0.15, B = 0.45, mean_v = 2.25, sd_v = 0.15, t0 = 0.2)

tnorm_prior <- BuildPrior(
    p0 = p0,
    p1 = rep(1, 5),
    lower = rep(0, 5),
    upper = rep(NA, 5),
    dist = rep("tnorm", 5),
    log_p = rep(FALSE, 5)
)


gamma_prior <- BuildPrior(
    p0 = p0,
    p1 = rep(0.1, 5),
    lower = rep(NA, 5),
    upper = rep(NA, 5),
    dist = rep("gamma", 5),
    log_p = rep(FALSE, 5)
)


lnorm_prior <- BuildPrior(
    p0 = p0,
    p1 = rep(0.1, 5),
    lower = rep(NA, 5),
    upper = rep(NA, 5),
    dist = rep("lnorm", 5),
    log_p = rep(FALSE, 5)
)

cauchy_prior <- BuildPrior(
    p0 = p0,
    p1 = rep(0.1, 5),
    lower = rep(NA, 5),
    upper = rep(NA, 5),
    dist = rep("cauchy", 5),
    log_p = rep(FALSE, 5)
)

unif_prior <- BuildPrior(
    p0 = p0,
    p1 = rep(0.1, 5),
    lower = rep(NA, 5),
    upper = rep(NA, 5),
    dist = rep("unif", 5),
    log_p = rep(FALSE, 5)
)

norm_prior <- BuildPrior(
    p0 = p0,
    p1 = rep(0.1, 5),
    lower = rep(NA, 5),
    upper = rep(NA, 5),
    dist = rep("norm", 5),
    log_p = rep(FALSE, 5)
)


print(plot_prior(tnorm_prior, font_size = 2.5, cex = 2.5))
print(plot_prior(gamma_prior, font_size = 2.5, cex = 2.5))
print(plot_prior(lnorm_prior, font_size = 2.5, cex = 2.5))
print(plot_prior(cauchy_prior, font_size = 2.5, cex = 2.5))
print(plot_prior(unif_prior, font_size = 2.5, cex = 2.5))
print(plot_prior(norm_prior, font_size = 2.5, cex = 2.5))



nparameter <- length(p0)
for (i in seq_len(nparameter)) {
    tnorm_prior[[i]]$log_p <- TRUE
    gamma_prior[[i]]$log_p <- TRUE
    lnorm_prior[[i]]$log_p <- TRUE

    cauchy_prior[[i]]$log_p <- TRUE
    unif_prior[[i]]$log_p <- TRUE
    norm_prior[[i]]$log_p <- TRUE
}
res <- print_prior(tnorm_prior)
res <- print_prior(gamma_prior)
res <- print_prior(lnorm_prior)
res <- print_prior(cauchy_prior)
res <- print_prior(unif_prior)
res <- print_prior(norm_prior)

set.seed(123)
parameters_r <- runif(nparameter, 0, 10)

res0 <- sumlogprior(p_prior_r = tnorm_prior, parameters_r = parameters_r)
res1 <- sumlogprior(p_prior_r = gamma_prior, parameters_r = parameters_r)
res2 <- sumlogprior(p_prior_r = lnorm_prior, parameters_r = parameters_r)
res3 <- sumlogprior(p_prior_r = cauchy_prior, parameters_r = parameters_r)
res4 <- sumlogprior(p_prior_r = unif_prior, parameters_r = parameters_r)
res5 <- sumlogprior(p_prior_r = norm_prior, parameters_r = parameters_r)

results <- c(res0, res1, res2, res3, res4, res5)

Prerequisites

R (>= 3.5.0), Rcpp (>= 1.0.7), RcppArmadillo (>= 0.10.7.5.0), ggdmcHeaders, lattice.

See DESCRIPTION for details

Installation

From CRAN:

install.packages("ggdmcPrior")