GOFmeasures: Goodness of fit measures
In ilapros/GOFmeas: Goodness of fit measures for regional flood frequency analysis

Description Usage Arguments Details Value Author(s) References See Also Examples

View source: R/HydroGOF.R

The main function to compute the Hosking and Wallis (1997) and Kjeldsen and Prosdocimi (2014) goodness of fit measures.

GOFmeasures(
  stations = NULL,
  lmom = NULL,
  n.amax = NULL,
  Nsim = 500,
  mcmom = NULL,
  type = c("HW_GOF", "KP_GOF"),
  plot = FALSE,
  conf.lev = 0.9,
  dist.type = "mahalanobis",
  ...
)

`stations`	a list of amax, with one list components per station (see `Details`)
`lmom`	the matrix of sample lmoments: each line corresponds to the lmoments of for each station (see `Details`)
`n.amax`	a vector with the length of years recorded in each station (see `Details`)
`Nsim`	the number of simulations used in the creation of mcmom, if `mcmom` is left null.
`mcmom`	an (Nsim*4) Matrix containing regional average L-moments (l1,l2,t3,t4) for the Nsim MC generated homogeneous regions. By default this is left `NULL` and the mcmom matrix is computed.
`type`	indicates which GOF measure should be calculated, default is both "HW" and "KP"
`plot`	logical vector indicating if the lmoment diagram with the ellipses intersecting the theoretical lines should be displayed
`conf.lev`	confidence level (1-alpha). If for a given distribution the distance between the bias-corrected tR vector and the theoretical L-skewness and L-kurtosis line is bigger than the qchisq(1-alpha, 2) value the distribution is not considered as an acceptable distribution.
`dist.type`	the type of distance to be used in the Kjeldsen and Prosdocimi measure to determine the distance between the bias-corrected regional parameters and the theoretical L-skewness and L-kurtosis line, default set to "mahalanobis", the other option ("geometric") computes the simple geometric distance
`...`	arguments to pass to `mahalanobis`

The function can take information on the annual maxima information either as a list of stations in stations or as the summarised lmom matrix (lmom) and information of the length of each series (n.amax).

An object of the GOFmeas class.

`GOFtable`	a table with the Goodness of fit results
`selectedDist`	the selected distribution which minimises the goodness of fit measures calculated
`mcmom`	the (Nsim*4) mcmom matrix of sample lmoments from the MCgenerated homogeneous region. The matrix can be reused if needed
`lmom`	the matrix of sample lmoments for the region under study
`rmom`	the weighted average regional lmoment. Weights are given according to the length of sample at each station
`n.amax`	a vector with the number of years recorded in each station

Thomas Kjeldsen and Ilaria Prosdocimi

Hosking, J. R. M., and J. R. Wallis (1997), Regional frequency analysis: an approach based on L-moments, Cambridge University Press.

Kjeldsen T. R. and Prosdocimi I. (2015) A bivariate extension of the Hosking and Wallis goodness-of-fit measure for regional distributions. Water Resources Research. DOI: 10.1002/2014WR015912

print.GOFmeas summary.GOFmeas

set.seed(546)
# generate a region with (mostly) GEV distribution
library(lmom)
stats <- list(quagev(runif(30,0,1),c(289, 45 ,-0.22)),
              quagev(runif(45,0,1),c(189, 40 ,-0.15)),
              quagev(runif(25,0,1),c(122, 10 ,-0.24)),
              quagev(runif(43,0,1),c( 59,  8 ,-0.18)),
              quagev(runif(32,0,1),c( 62, 10 ,-0.21)),
              quagev(runif(28,0,1),c( 91,  9 ,-0.25)),
              quaglo(runif(27,0,1),c(202, 25 ,-0.17)))
tt <- GOFmeasures(stats)
tt
summary(tt)