R/ts.non.seas.model.r

In okgreece/TimeSeries.OBeu: Time Series Analysis 'OpenBudgets.eu'

#' @title 
#' Model fit of non seasonal time series
#'
#' @description
#' Model fit of non seasonal time series
#'
#' @usage ts.non.seas.model(tsdata, x.ord = NULL, tojson = FALSE)
#' 
#' @param tsdata The input univariate non seasonal time series data
#' @param x.ord An integer vector of length 3 specifying the order of the Arima model
#' @param tojson If TRUE the results are returned in json format, default returns a list

#' @details 
#' Model fit of non seasonal time series using arima models of non seasonal time series data.
#' The model with the lowest AIC value is selected for forecasts.
#' 
#' @return 
#' A list with the following components:
#' \itemize{
#' \item model.summary: 
#' \itemize{
#' \item ts_model The summary model details returned as Arima object for internal use in ts.analysis function}
#'
#' \item model:
#' \itemize{
#'  \item ts_model: 
#'  \item arima.order The Arima order
#'  \item arima.coef A vector of AR, MA and regression coefficients
#'  \item arima.coef.se The standard error of the coefficients }
#' 
#' \item residuals: The residuals of the model (fitted innovations)
#' 
#' \item compare:
#' \itemize{
#'  \item variance.coef The matrix of the estimated variance of the coefficients
#'  \item resid.variance The MLE of the innovations variance
#'  \item not.used.obs The number of not used observations for the fitting
#'  \item used.obs the number of used observations for the fitting
#'  \item loglik The maximized log-likelihood (of the differenced data), or the approximation to it used
#'  \item aic The AIC value corresponding to the log-likelihood
#'  \item bic The BIC value corresponding to the log-likelihood
#'  \item aicc The second-order Akaike Information Criterion corresponding to the log-likelihood}}
#'  
#' @author Kleanthis Koupidis
#' 
#' @seealso \code{\link{ts.analysis}}, \code{\link[forecast]{Arima}}
#' 
#' @examples
#' ts.non.seas.model(Athens_draft_ts)
#' 
#' @rdname ts.non.seas.model
#' 
#' @export
#' 

ts.non.seas.model <- function(tsdata, x.ord = NULL, tojson = FALSE){
  
  #arima obeu
  arima.obeu <- function(tsdata, x) {
    tryCatch(forecast::Arima(tsdata, order = c(x, 1, 1)),
             warning = function(w) {#print(paste("next order", x)); 
               NULL},
             error = function(e) {#print(paste("next order", x)); 
               NULL}) 
  }
  #aic.obeu
  aic.obeu <- function(aic, x) {
    tryCatch(c(aic=modelss[[x]]$aic,order = c(x, 1, 1)),
             warning = function(w) {#print(paste("next order", x)); 
               NULL},
             error = function(e) {#print(paste("next order", x)); 
               NULL}) 
  }
  
  if (is.null(x.ord) == FALSE | 
      all(x.ord == c(0, 0, 0)) == FALSE) {
    x <- x.ord
  } else  {
    #Selection of the appropriate model
    aiccc <- list()
    modelss <- list()
    
    for(i in 1:8) {
      
      modelss[[i]] <- arima.obeu(tsdata,i)
      aiccc[[i]] <- aic.obeu(modelss[[i]]$aic,i)
    }
    
    df <- data.frame(matrix(unlist(aiccc), ncol = 4, byrow = TRUE))
    colnames(df) <- c("aic", "ar", "diff", "ma")
    mindf <- df[order(df$aic),]
    mindf <- mindf[3,]
    x <- c(mindf$ar, mindf$diff, mindf$ma)
  } 
  # Fit the appropriate model
  ts_model <- forecast::Arima(y = tsdata, order = x)
  model.summary <- ts_model
  
  model <- list(	# Model
    arima.order = ts_model$arma,
    arima.coef = ts_model$coef,
    arima.coef.se = round(sqrt(diag(ts_model$var.coef)),digits=4))
  
  
  residuals_fitted <- list(
    residuals = ts_model$residuals,
    fitted = stats::fitted(ts_model),
    time = stats::time(tsdata),
    line = 0)
  
  compare <- list(
    resid.variance = ts_model$sigma2,
    variance.coef = ts_model$var.coef,
    not.used.obs = ts_model$n.cond,
    used.obs = ts_model$nobs,
    loglik = ts_model$loglik,
    aic = ts_model$aic,
    bic = ts_model$bic,
    aicc = ts_model$aicc)
  
  model.details <- list(
    model.summary = model.summary, 
    model = model,
    residuals_fitted = residuals_fitted,
    compare = compare)
  
  if (tojson == TRUE) {
    model.details <- jsonlite::toJSON(model.details)
  }
  
  return(model.details)
}