R/RcppExports.R
In reykp/BEDr: What the Package Does in One 'Title Case' Line
# Generated by using Rcpp::compileAttributes() -> do not edit by hand
# Generator token: 10BE3573-1514-4C36-9D1C-5A225CD40393
#' @title duos (Distribution of Uniform Order Statistics)
#' @description
#' Runs Gibbs sampling algorithm to perform Bayesian density estimation through \code{duos}.
#'
#' @usage
#'
#' duos(y, k = [n/50] + 3, N = 20000, alpha = 1, scale_l = 0.00001, scale_u = 0.00001, start = NA)
#'
#' @name duos
#' @param y A numeric vector. \code{duos} estimates the density on this data.
#' @param k The number of cut-points for the density estimate. There is a default based on the size of \code{y} (see details).
#' @param N The number of iterations to run in the algorithm. The default is 20,000.
#' @param alpha The parameter values for the Dirichlet prior. The parameter is constant and set to 1 by default.
#' @param scale_l A value >= 0 controlling the scaling based on the minimum data value. The default is 0.00001 (see details).
#' @param scale_u A value >= 0 controlling the scaling based on the maximum data value. The default is 0.00001 (see details).
#' @param start A vector of starting values of length k. This is useful for creating output to use in Gelman and Rubin's convergence diagnostics (see examples). If NA, reasonable starting values are selected by the algorithm.
#' @examples
#'
#' @export
#'
#' @details
#'
#' The density being estimated takes the form below:
#'
#' \eqn{f(x) =}
#' \deqn{(\pi_1) / (\gamma_1) , 0 \le x < \gamma_1}
#' \deqn{(\pi_2) / (\gamma_2-\gamma_1) , \gamma_1 \le x < \gamma_2}
#' \deqn{(\pi_3) / (\gamma_3-\gamma_2) , \gamma_2 \le x < \gamma_3}
#' \deqn{... , ... \le x < ...}
#' \deqn{(\pi_k) / (\gamma_k-\gamma_{k-1}) , \gamma_{k-1} \le x < \gamma_k}
#' \deqn{(\pi_{k+1}) / (1-\gamma_k) , \gamma_k \le x < 1}
#'
#' where \eqn{\gamma_1 < \gamma_2 < ... < \gamma_k is in (0,1) and \pi_1 + \pi_2 + ... + \pi_{k+1} = 1}
#'
#' The prior on the \eqn{\gamma's} is: draw k values from $unif(0,1)$ and order: $0 < \gamma_1 < \gamma_2 < ... < \gamma_{k-1} < \gamma_k < 1$.
#' The prior on the \eqn{\pi's} is Dirichlet(\eqn{\alpha}).
#'
#' This density operates on data between 0 and 1. Thus, if the input is not between 0 and 1, it is standardized. The formula for scaling is below:
#'
#' \deqn{(y-(min(y)-scale_l))/(max(y)+scale_u-(min(y)-scale_l))}.
#'
#' Values of 0 for the scale parameters indicates the density will only be the edges of the data in \code{y}. The default for \code{scale_l} and \code{scale_u} is 0.0001.
#'
#' The recommended number of cut-points starts at 3, and then for each increment of 50, adds a cut-point.
#'
#' Default: k = round(n/50)+3
#'
#' @return
#'
#' \code{duos} returns a list of density estimate results.
#'
#' \item{\code{C}}{A matrix containing the posterior draws for the cut-point parameters. The number of columns is \code{k}, and the number of rows is \code{N}.}
#' \item{\code{P}}{A matrix containing the posterior draws for the bin proportion parameters. The number of columns is \code{k}+1, and the number of rows is \code{N}.}
#' \item{\code{y}}{A vector containing the data introduced to \code{duos} for density estimation.}
#' \item{\code{k}}{The number of cut-points.}
#' \item{\code{alpha}}{The parameter for the prior on the bin proportions.}
#' \item{\code{scale_l}}{The scaling parameter for the lower end of the data.}
#' \item{\code{scale_u}}{The scaling parameter for the upper end of the data.}
#'
#' @examples
#' ## --------------------------------------------------------------------------------
#' ## Beta Distribution
#' ## --------------------------------------------------------------------------------
#'
#' # Run 'duos' on data sampled from a Beta(5, 1) distribution with 100 data points.
#' # All defaults are used.
#' y <- rbeta(100, 5, 1)
#' duos_beta <- duos(y)
#'
#' #Examine estimate of PDF
#' duos_plot(duos_beta, type = "pdf", data = TRUE)
#'
#' #Examine estimate of CDF with credible intervals
#' duos_plot(duos_beta, type = "cdf")
#'
#' #Find probability of being less than 0.1
#' duos_cdf(c(.1), duos_beta)$cdf
#'
#' ## --------------------------------------------------------------------------------
#' ## Claw Distribution
#' ## --------------------------------------------------------------------------------
#' # Run 'duos' on data sampled from a 'Claw' distribution with 200 data points.
#' u <- runif(200)
#'
#' # Variable to store the samples from the mixture distribution
#' y <- rep(NA, 200)
#' # Sampling from the mixture
#' for(i in 1:200){
#' if(u[i] < 0.5){
#' y[i] <- rnorm(1, 0, 1)
#' }else if(u[i] < 0.6){
#' y[i] <- rnorm(1, -1, 0.1)
#' }else if (u[i] < 0.7){
#' y[i] <- rnorm(1, -0.5, 0.1)
#' }else if (u[i] < 0.8){
#' y[i] <- rnorm(1, 0, 0.1)
#' }else if (u[i] <- 0.9){
#' y[i] <- rnorm(1, 0.5, 0.1)
#' }else{
#' y[i] <- rnorm(1, 1, 0.1)
#' }
#' }
#'
#' # Choose 12 cutpoints
#' duos_claw <- duos(y = y, k = 12, N = 20000)
#'
#' # Examine estimate of PDF
#' duos_plot(duos_claw, type = "pdf", data = TRUE)
#'
#' # Examine estimate of CDF
#' duos_plot(duos_claw, type = "cdf")
#'
#' # Find probability of being greater than 90
#' 1-duos_cdf(c(0.90), duos_claw)$cdf
#'
#' ## --------------------------------------------------------------------------------
#' ## Normal Distribution
#' ## --------------------------------------------------------------------------------
#'
#' # Run 'duos' on data sampled from a Normal(0, 1) distribution with 50 data points.
#' y <- rnorm(50, 0, 1)
#' # Run for 15,000 iterations and change the 'scale' values to 1.5*sd(y)
#' duos_norm <- duos(y = y, scale_l = 1.5*sd(y), scale_u = 1.5*sd(y), N = 15000)
#'
#' # Examine estimate of PDF
#' duos_plot(duos_norm, type = "pdf", data = TRUE)
#'
#' # Examine estimate of CDF
#' duos_plot(duos_norm, type = "cdf", cri = TRUE)
#'
#' # Find probability of being greater than 2.5
#' 1-duos_cdf(c(2.5), duos_norm)$cdf
#'
#' ## --------------------------------------------------------------------------------
#' ## Uniform Distribution
#' ## --------------------------------------------------------------------------------
#'
#' # Below is an example of using the 'start' variable to calculate the Gelman and Rubin diagnostic
#'
#' # First run 'duos' on data sampled from a Uniform(0, 1) distribution with 50 data points.
#' y <- runif(50)
#'
#' # Use 4 cutpoints
#' # Create three runs with diverse starting values
#' duos_unif1 <- duos(y = y, k = 4, N = 20000, start = runif(4, 0, 1/3))
#' duos_unif2 <- duos(y = y, k = 4, N = 20000, start = runif(4, 1/3, 2/3))
#' duos_unif3 <- duos(y = y, k = 4, N = 20000, start = runif(4, 2/3, 1))
#'
#' # Load package coda
#' library(coda)
#'
#' # Turn matrices of cut-points into mcmc objects
#' C1 <- mcmc(duos_unif1$C)
#' C2 <- mcmc(duos_unif2$C)
#' C3 <- mcmc(duos_unif3$C)
#'
#' # Turn into mcmc list
#' all_chains <- mcmc.list(C1, C2, C3)
#'
#' #Get Gelman and Rubin diagnostic
#' gelman.diag(all_chains)[[1]][,1]
#'
#' #Examine estimate of PDF's from each run
#' duos_plot(duos_unif1, type="pdf", data=TRUE)
#' duos_plot(duos_unif2, type="pdf", data=TRUE)
#' duos_plot(duos_unif3, type="pdf", data=TRUE)
NULL
duos <- function(y, k = NA_real_, N = 20000, alpha = 1L, scale_l = 0.00001, scale_u = 0.00001, start = NA_real_) {
.Call('_biRd_duos', PACKAGE = 'biRd', y, k, N, alpha, scale_l, scale_u, start)
}
reykp/BEDr documentation built on May 28, 2019, 8:40 a.m.
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
# Generated by using Rcpp::compileAttributes() -> do not edit by hand
# Generator token: 10BE3573-1514-4C36-9D1C-5A225CD40393
#' @title duos (Distribution of Uniform Order Statistics)
#' @description
#' Runs Gibbs sampling algorithm to perform Bayesian density estimation through \code{duos}.
#'
#' @usage
#'
#' duos(y, k = [n/50] + 3, N = 20000, alpha = 1, scale_l = 0.00001, scale_u = 0.00001, start = NA)
#'
#' @name duos
#' @param y A numeric vector. \code{duos} estimates the density on this data.
#' @param k The number of cut-points for the density estimate. There is a default based on the size of \code{y} (see details).
#' @param N The number of iterations to run in the algorithm. The default is 20,000.
#' @param alpha The parameter values for the Dirichlet prior. The parameter is constant and set to 1 by default.
#' @param scale_l A value >= 0 controlling the scaling based on the minimum data value. The default is 0.00001 (see details).
#' @param scale_u A value >= 0 controlling the scaling based on the maximum data value. The default is 0.00001 (see details).
#' @param start A vector of starting values of length k. This is useful for creating output to use in Gelman and Rubin's convergence diagnostics (see examples). If NA, reasonable starting values are selected by the algorithm.
#' @examples
#'
#' @export
#'
#' @details
#'
#' The density being estimated takes the form below:
#'
#' \eqn{f(x) =}
#' \deqn{(\pi_1) / (\gamma_1) , 0 \le x < \gamma_1}
#' \deqn{(\pi_2) / (\gamma_2-\gamma_1) , \gamma_1 \le x < \gamma_2}
#' \deqn{(\pi_3) / (\gamma_3-\gamma_2) , \gamma_2 \le x < \gamma_3}
#' \deqn{... , ... \le x < ...}
#' \deqn{(\pi_k) / (\gamma_k-\gamma_{k-1}) , \gamma_{k-1} \le x < \gamma_k}
#' \deqn{(\pi_{k+1}) / (1-\gamma_k) , \gamma_k \le x < 1}
#'
#' where \eqn{\gamma_1 < \gamma_2 < ... < \gamma_k is in (0,1) and \pi_1 + \pi_2 + ... + \pi_{k+1} = 1}
#'
#' The prior on the \eqn{\gamma's} is: draw k values from $unif(0,1)$ and order: $0 < \gamma_1 < \gamma_2 < ... < \gamma_{k-1} < \gamma_k < 1$.
#' The prior on the \eqn{\pi's} is Dirichlet(\eqn{\alpha}).
#'
#' This density operates on data between 0 and 1. Thus, if the input is not between 0 and 1, it is standardized. The formula for scaling is below:
#'
#' \deqn{(y-(min(y)-scale_l))/(max(y)+scale_u-(min(y)-scale_l))}.
#'
#' Values of 0 for the scale parameters indicates the density will only be the edges of the data in \code{y}. The default for \code{scale_l} and \code{scale_u} is 0.0001.
#'
#' The recommended number of cut-points starts at 3, and then for each increment of 50, adds a cut-point.
#'
#' Default: k = round(n/50)+3
#'
#' @return
#'
#' \code{duos} returns a list of density estimate results.
#'
#' \item{\code{C}}{A matrix containing the posterior draws for the cut-point parameters. The number of columns is \code{k}, and the number of rows is \code{N}.}
#' \item{\code{P}}{A matrix containing the posterior draws for the bin proportion parameters. The number of columns is \code{k}+1, and the number of rows is \code{N}.}
#' \item{\code{y}}{A vector containing the data introduced to \code{duos} for density estimation.}
#' \item{\code{k}}{The number of cut-points.}
#' \item{\code{alpha}}{The parameter for the prior on the bin proportions.}
#' \item{\code{scale_l}}{The scaling parameter for the lower end of the data.}
#' \item{\code{scale_u}}{The scaling parameter for the upper end of the data.}
#'
#' @examples
#' ## --------------------------------------------------------------------------------
#' ## Beta Distribution
#' ## --------------------------------------------------------------------------------
#'
#' # Run 'duos' on data sampled from a Beta(5, 1) distribution with 100 data points.
#' # All defaults are used.
#' y <- rbeta(100, 5, 1)
#' duos_beta <- duos(y)
#'
#' #Examine estimate of PDF
#' duos_plot(duos_beta, type = "pdf", data = TRUE)
#'
#' #Examine estimate of CDF with credible intervals
#' duos_plot(duos_beta, type = "cdf")
#'
#' #Find probability of being less than 0.1
#' duos_cdf(c(.1), duos_beta)$cdf
#'
#' ## --------------------------------------------------------------------------------
#' ## Claw Distribution
#' ## --------------------------------------------------------------------------------
#' # Run 'duos' on data sampled from a 'Claw' distribution with 200 data points.
#' u <- runif(200)
#'
#' # Variable to store the samples from the mixture distribution
#' y <- rep(NA, 200)
#' # Sampling from the mixture
#' for(i in 1:200){
#' if(u[i] < 0.5){
#' y[i] <- rnorm(1, 0, 1)
#' }else if(u[i] < 0.6){
#' y[i] <- rnorm(1, -1, 0.1)
#' }else if (u[i] < 0.7){
#' y[i] <- rnorm(1, -0.5, 0.1)
#' }else if (u[i] < 0.8){
#' y[i] <- rnorm(1, 0, 0.1)
#' }else if (u[i] <- 0.9){
#' y[i] <- rnorm(1, 0.5, 0.1)
#' }else{
#' y[i] <- rnorm(1, 1, 0.1)
#' }
#' }
#'
#' # Choose 12 cutpoints
#' duos_claw <- duos(y = y, k = 12, N = 20000)
#'
#' # Examine estimate of PDF
#' duos_plot(duos_claw, type = "pdf", data = TRUE)
#'
#' # Examine estimate of CDF
#' duos_plot(duos_claw, type = "cdf")
#'
#' # Find probability of being greater than 90
#' 1-duos_cdf(c(0.90), duos_claw)$cdf
#'
#' ## --------------------------------------------------------------------------------
#' ## Normal Distribution
#' ## --------------------------------------------------------------------------------
#'
#' # Run 'duos' on data sampled from a Normal(0, 1) distribution with 50 data points.
#' y <- rnorm(50, 0, 1)
#' # Run for 15,000 iterations and change the 'scale' values to 1.5*sd(y)
#' duos_norm <- duos(y = y, scale_l = 1.5*sd(y), scale_u = 1.5*sd(y), N = 15000)
#'
#' # Examine estimate of PDF
#' duos_plot(duos_norm, type = "pdf", data = TRUE)
#'
#' # Examine estimate of CDF
#' duos_plot(duos_norm, type = "cdf", cri = TRUE)
#'
#' # Find probability of being greater than 2.5
#' 1-duos_cdf(c(2.5), duos_norm)$cdf
#'
#' ## --------------------------------------------------------------------------------
#' ## Uniform Distribution
#' ## --------------------------------------------------------------------------------
#'
#' # Below is an example of using the 'start' variable to calculate the Gelman and Rubin diagnostic
#'
#' # First run 'duos' on data sampled from a Uniform(0, 1) distribution with 50 data points.
#' y <- runif(50)
#'
#' # Use 4 cutpoints
#' # Create three runs with diverse starting values
#' duos_unif1 <- duos(y = y, k = 4, N = 20000, start = runif(4, 0, 1/3))
#' duos_unif2 <- duos(y = y, k = 4, N = 20000, start = runif(4, 1/3, 2/3))
#' duos_unif3 <- duos(y = y, k = 4, N = 20000, start = runif(4, 2/3, 1))
#'
#' # Load package coda
#' library(coda)
#'
#' # Turn matrices of cut-points into mcmc objects
#' C1 <- mcmc(duos_unif1$C)
#' C2 <- mcmc(duos_unif2$C)
#' C3 <- mcmc(duos_unif3$C)
#'
#' # Turn into mcmc list
#' all_chains <- mcmc.list(C1, C2, C3)
#'
#' #Get Gelman and Rubin diagnostic
#' gelman.diag(all_chains)[[1]][,1]
#'
#' #Examine estimate of PDF's from each run
#' duos_plot(duos_unif1, type="pdf", data=TRUE)
#' duos_plot(duos_unif2, type="pdf", data=TRUE)
#' duos_plot(duos_unif3, type="pdf", data=TRUE)
NULL
duos <- function(y, k = NA_real_, N = 20000, alpha = 1L, scale_l = 0.00001, scale_u = 0.00001, start = NA_real_) {
.Call('_biRd_duos', PACKAGE = 'biRd', y, k, N, alpha, scale_l, scale_u, start)
}
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Embedding an R snippet on your website
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.