R/degree.distr.fitter.R

In wdelva/RSimpactHelp: Pre- and Post-simulation helper functions for RSimpact Cyan

#' Fitting parametric distributions to the degree data
#'
#' The user fits various parametric distributions to the degree data and
#' compare their goodness of fit.To choose on the best fitting distribution,
#' a table is returned giving the empirical estimate of the three considered
#' goodness-of-fit statistics.
#'
#' @param dataframe The dataframe that is produced by \code{\link{degree.df.maker}}
#'
#'
#' @return A table of the empirical estimate of the three considered goodness-of-fit statistics.
#'
# @examples
# load(dataframe)
#bestfit <- degree.distr.fitter(dataframe = degree.df)
#'
#' @import fitdistrplus
#' @export


# Fitting the parametric distributions to the Degree data by
# maximum likelihood estimation and comparison of goodness of fit statistics.

degree.distr.fitter <- function(dataframe = degree.df){
  # Normal distribution
  (fitn<-fitdistrplus::fitdist(dataframe$Degree, "norm"))
  #gofstat(fitn)

  # Poisson distribution
  (fitpois<-fitdistrplus::fitdist(dataframe$Degree, "pois"))
  #gofstat(fitpois)

  # Negative Bionomial distribution
  (fitnb<-fitdistrplus::fitdist(dataframe$Degree, "nbinom"))
  #gofstat(fitnb)

  # Weibull distribution
  (fitw<-fitdistrplus::fitdist(dataframe$Degree, "weibull"))
  #gofstat(fitw)

  # Gamma distribution
  (fitg<-fitdistrplus::fitdist(dataframe$Degree, "gamma"))
  #gofstat(fitg)

  # Log-normal distribution
  (fitln<-fitdistrplus::fitdist(dataframe$Degree, "lnorm"))
  #gofstat(fitln)

  # Beta distribution

  # (fitb<-fitdist(dataframe$degree, "beta"))
  # gofstat(fitb) # values must be in [0-1]

  fit.list <- list(fitn,fitpois,fitnb,fitw,fitg,fitln)


  (fit<-fitdistrplus::gofstat(f = fit.list,
                              chisqbreaks = c(0,1,4,10,1000)))

  bestfit.logic <- as.logical(fit$aic == min(fit$aic))
  bestfit.place <- min(which(bestfit.logic == TRUE))

  return(fit.list[[bestfit.place]])
}