R/OptPartitioning2D.R

In lpishchagina/OptPartitioning2D: Optimal Partitioning and PELT algorithm

#' @title GenData2D
#'  
#' @description Generation of data of dimension 2 with a given values of means and changepoints
#' 
#' @param n number of data point.
#' @param changepoints a vector of increasing changepoint indices (last element is always n!).
#' @param means1 vector of successive means1 for data1.
#' @param means2 vector of successive means2 for data2.
#' @param noise1 standard deviation of an additional normal noise (for data1).
#' @param noise2 standard deviation of an additional normal noise (for data2).
#'  
#' @return matrix of data of dimension 2 x n with a given values of means by the segmentation.
#'  
#' @examples
#' GenData2D (n = 10, changepoints = c(2, 4, 6, 8, 10), means1 = c (0, 1, 0, 1, 0), means2 = (1, 2, 3, 4, 5), noise1 = 1, noise2 = 1)

GenData2D <- function(n, changepoints, means1, means2, noise1 = 1, noise2 = 1)
{
  #---stop---#
  if (n!= changepoints[length(changepoints)]) {stop('last element of changepoints is always n')}
  
  if(!is.numeric(changepoints)) {stop('changepoints are not all numeric')}
  if(is.unsorted(changepoints)) {stop('changepoints should be an increasing vector')}
  
  if(!is.numeric(means1)){stop('means1 are not all numeric')}
  if(!is.numeric(means2)){stop('means2 are not all numeric')}
  if( length(means1) != (length(means2)) ) {stop('means1 and means2 vectors are of different size')}
  if( length(changepoints) != length(means1) ) {stop('changepoints and means1 vectors are of different size')}
  if( length(changepoints) != length(means2) ) {stop('changepoints and means2 vectors are of different size')}
  
  if(!is.double(noise1)) {stop('noise1 is not a double')}
  if(noise1 < 0) {stop('noise1 must be non-negative')}
  if(!is.double(noise2)) {stop('noise2 is not a double')}
  if(noise2 < 0) {stop('noise2 must be non-negative')}
  #---function---#	
  data <- matrix(0,2,n)
  InttT<- diff(c(0,changepoints))
  # rnorm(mu,noise) = mu + rnorm(0,noise)
  data[1,] <- rep(means1, InttT) + rnorm(n, 0, noise1)
  data[2,] <- rep(means2, InttT) + rnorm(n, 0, noise2)
  return(data)
}

#' @title PlotOptPart2D
#'  
#' @description Plot of data with a  values of means and changepoints.
#' 
#' @param data matrix of data of dimension 2 x n.
#' @param n number of data point.
#' 
#' @param changepoints a vector of increasing changepoint indices (last element is always n!).
#' @param means1 vector of successive means1 for data1.
#' @param means2 vector of successive means2 for data2.

#' @return Plot of data with a given values of means and changepoints.
#'  
#' @examples
#' PlotOptPart2D( data = matrix(c(1,0,1,0,1,0,1,0,2,3,2,3,2,3,2,3), nrow = 2), changepoints = c(5, 8), means1 = c(1, 2), means2 = c(0,3))

PlotOptPart2D <- function (data, changepoints, means1, means2)
{
  #---stop---#
  if (length(data[1,])!= changepoints[length(changepoints)]) {stop('last element of changepoints is always n')}
  
  if(!is.numeric(data)) {stop('data values are not all numeric')}
  if(is.unsorted(changepoints)) {stop('changepoints should be an increasing vector')}
  
  if(!is.numeric(means1)){stop('means1 are not all numeric')}
  if(!is.numeric(means2)){stop('means2 are not all numeric')}
  if( length(means1) != (length(means2)) ) {stop('means1 and means2 vectors are of different size')}
  if( length(changepoints) != length(means1) ) {stop('changepoints and means1 vectors are of different size')}
  if( length(changepoints) != length(means2) ) {stop('changepoints and means2 vectors are of different size')}
  
  #---function---#
  tau <- c(0, changepoints)
  par(mfrow = rbind(2,1))
  plot(c(1:length(data[1,])), data[1,], main = "Optimal Partitioning Y1", xlab = "time", ylab = "Y1", col = "blue")
  abline(v = changepoints, lty = 2, col = "grey")
  i<-1
  while(i != length(tau))
  {
    lines(c(tau[i],tau[i+1]), c(means1[i],means1[i]), type = "l", col = "red")
    i<-i+1
  }
  plot(c(1:length(data[2,])), data[2,], main = "Optimal Partitioning Y2 ", xlab = "time", ylab = "Y2", col = "red")
  abline(v = changepoints, lty = 2, col = "grey")
  i<-1
  while(i != length(tau))
  {
    lines(c(tau[i],tau[i+1]), c(means2[i],means2[i]), type = "l", col = "blue")
    i<-i+1
  }
}