R/CE.NB.R

In breakpoint: An R Package for Multiple Break-Point Detection via the Cross-Entropy Method

CE.NB <-
function(data, Nmax=10, eps=0.01, rho=0.05, M=200, h=5, a=0.8, b=0.8, distyp = 1, penalty = "BIC", parallel=FALSE){
  
  if(is.data.frame(data) == "FALSE" | is.null(dim(data)[2])) {
    print("Error in data: dataframe only")                                                           
  } else if(dim(data)[2] != 1) {
    print("Error in data: single column dataframe only")                               
  } else { 
    
    if(distyp == 1 & penalty == "BIC"){
      Melite <- M*rho
      L <- length(data[, 1])
      L0 <- 1      
      k <- seq(0, Nmax, 1)
      r <- suppressWarnings(try(fitdistr(data[, 1], "negative binomial")[[1]][[1]], silent = T)) 
      
      if(parallel == TRUE & .Platform$OS.type == "windows"){
          cl <- makeCluster(parallel::detectCores(), type = "SOCK") 
          clusterExport(cl, c("ce.4betaNB.BIC", "betarand", "fun.alpha", "fun.beta", "BICnb", "llhoodnb", "logliknb"), envir = environment())
          clusterExport(cl, c("data", "rho", "M", "h", "eps", "Melite", "L", "L0", "a", "r"), envir = environment())  
          registerDoParallel(cl)
          sim <- foreach(k = k, .errorhandling = c('pass')) %dopar% ce.4betaNB.BIC(k, data, h, L0, L, M, Melite, eps, a, r)
          stopCluster(cl)   
          
          
        } else if (parallel == TRUE & .Platform$OS.type == "unix"){
            registerDoParallel(parallel::detectCores()) 
            sim <- foreach(k = k, .errorhandling = c('pass')) %dopar% ce.4betaNB.BIC(k, data, h, L0, L, M, Melite, eps, a, r)
          
        } else { 
          sim <- foreach(k = k, .errorhandling = c('pass')) %do% ce.4betaNB.BIC(k, data, h, L0, L, M, Melite, eps, a, r)
          }
      
      Bic.summary <- sapply(sim, "[[", 2)
      opt.loci <- which(Bic.summary == min(Bic.summary))
      
      loci.BIC <- sim[[opt.loci]]$loci
      
      if(length(loci.BIC) >= 3) {
        return(list("No.BPs" = length(loci.BIC) - 2,"BP.Loc" = loci.BIC[2 : (length(loci.BIC) - 1)], "BIC" = sim[[opt.loci]]$BIC, "ll" = sim[[opt.loci]]$LogLike))
      } else {
        return(paste("No Break-Points are Estimated")) 
      }
    
    } else if(distyp == 2 & penalty == "BIC"){
      
      Melite <- M*rho
      L <- length(data[, 1])
      L0 <- 1      
      k <- seq(0, Nmax, 1)
      r <- suppressWarnings(try(fitdistr(data[, 1], "negative binomial")[[1]][[1]], silent = T)) 
      
      if(parallel == TRUE & .Platform$OS.type == "windows"){
          cl <- makeCluster(parallel::detectCores(), type = "SOCK") 
          clusterExport(cl, c("ce.simNormalNB.BIC", "normrand", "BICnb", "llhoodnb", "logliknb"), envir = environment())
          clusterExport(cl, c("data", "rho", "M", "h", "eps", "Melite", "L", "L0", "a", "b", "r"), envir=environment())  
          registerDoParallel(cl)
          sim <- foreach(k = k, .errorhandling = c('pass')) %dopar% ce.simNormalNB.BIC(k, data, h, L0, L, M, Melite, eps, a, b, r)
          stopCluster(cl)   
          
          
        } else if (parallel == TRUE & .Platform$OS.type == "unix"){
          registerDoParallel(parallel::detectCores()) 
          sim <- foreach(k = k, .errorhandling = c('pass')) %dopar% ce.simNormalNB.BIC(k, data, h, L0, L, M, Melite, eps, a, b, r)
            
        
        } else {
          sim <- foreach(k = k, .errorhandling = c('pass')) %do% ce.simNormalNB.BIC(k, data, h, L0, L, M, Melite, eps, a, b, r)
        }
      
      Bic.summary <- sapply(sim, "[[", 2)
      opt.loci <- which(Bic.summary == min(Bic.summary))
      loci.BIC <- sim[[opt.loci]]$loci
      
      if(length(loci.BIC) >= 3) {
        return(list("No.BPs" = length(loci.BIC) - 2,"BP.Loc" = loci.BIC[2 : (length(loci.BIC) - 1)], "BIC" = sim[[opt.loci]]$BIC, "ll" = sim[[opt.loci]]$LogLike))
      } else {
        return(paste("No Break-Points are Estimated")) 
      }          
    
    } else if(distyp == 1 & penalty == "AIC"){
      
      Melite <- M*rho
      L <- length(data[, 1])
      L0 <- 1      
      k <- seq(0, Nmax, 1)
      r <- suppressWarnings(try(fitdistr(data[, 1], "negative binomial")[[1]][[1]], silent = T)) 
      
      if(parallel == TRUE & .Platform$OS.type == "windows"){
          cl <- makeCluster(parallel::detectCores(), type = "SOCK") 
          clusterExport(cl, c("ce.4betaNB.AIC", "betarand", "fun.alpha", "fun.beta", "AICnb", "llhoodnb", "logliknb"), envir = environment())
          clusterExport(cl, c("data", "rho", "M", "h", "eps", "Melite", "L", "L0", "a", "r"), envir = environment())  
          registerDoParallel(cl)
          sim <- foreach(k = k, .errorhandling = c('pass')) %dopar% ce.4betaNB.AIC(k, data, h, L0, L, M, Melite, eps, a, r)
          stopCluster(cl)   
        
        
      } else if (parallel == TRUE & .Platform$OS.type == "unix"){
          registerDoParallel(parallel::detectCores()) 
          sim <- foreach(k = k, .errorhandling = c('pass')) %dopar% ce.4betaNB.AIC(k, data, h, L0, L, M, Melite, eps, a, r)
        
        
      } else {
        sim <- foreach(k = k, .errorhandling = c('pass')) %do% ce.4betaNB.AIC(k, data, h, L0, L, M, Melite, eps, a, r)
      }
      
      Aic.summary <- sapply(sim, "[[", 2)
      opt.loci <- which(Aic.summary == min(Aic.summary))
      loci.AIC <- sim[[opt.loci]]$loci
      
      if(length(loci.AIC) >= 3) {
        return(list("No.BPs" = length(loci.AIC) - 2,"BP.Loc" = loci.AIC[2 : (length(loci.AIC) - 1)], "AIC" = sim[[opt.loci]]$AIC, "ll" = sim[[opt.loci]]$LogLike))
      } else {
        return(paste("No Break-Points are Estimated")) 
      }          
      
    } else if(distyp == 2 & penalty == "AIC"){
      
      Melite <- M*rho
      L <- length(data[, 1])
      L0 <- 1      
      k <- seq(0, Nmax, 1)
      r <- suppressWarnings(try(fitdistr(data[, 1], "negative binomial")[[1]][[1]], silent = T)) 
      
      if(parallel == TRUE & .Platform$OS.type == "windows"){
          cl <- makeCluster(parallel::detectCores(), type = "SOCK") 
          clusterExport(cl, c("ce.simNormalNB.AIC", "normrand", "BICnb", "llhoodnb", "logliknb"), envir = environment())
          clusterExport(cl, c("data", "rho", "M", "h", "eps", "Melite", "L", "L0", "a", "b", "r"), envir=environment())  
          registerDoParallel(cl)
          sim <- foreach(k = k, .errorhandling = c('pass')) %dopar% ce.simNormalNB.AIC(k, data, h, L0, L, M, Melite, eps, a, b, r)
          stopCluster(cl)   
        
        
      } else if (parallel == TRUE & .Platform$OS.type == "unix"){
          registerDoParallel(parallel::detectCores()) 
          sim <- foreach(k = k, .errorhandling = c('pass')) %dopar% ce.simNormalNB.AIC(k, data, h, L0, L, M, Melite, eps, a, b, r)
        
      } else {
        sim <- foreach(k = k, .errorhandling = c('pass')) %do% ce.simNormalNB.AIC(k, data, h, L0, L, M, Melite, eps, a, b, r)
      }
      
      Aic.summary <- sapply(sim, "[[", 2)
      opt.loci <- which(Aic.summary == min(Aic.summary))
      loci.AIC <- sim[[opt.loci]]$loci
      
      if(length(loci.AIC) >= 3) {
        return(list("No.BPs" = length(loci.AIC) - 2,"BP.Loc" = loci.AIC[2 : (length(loci.AIC) - 1)], "AIC" = sim[[opt.loci]]$AIC, "ll" = sim[[opt.loci]]$LogLike))
      } else {
        return(paste("No Break-Points are Estimated")) 
      }          
    }
  }
}