tests/testthat/test-makeReferenceMarginalDistribution.R

In texmex: Statistical Modelling of Extreme Values

context("makeReferenceMarginalDistribution")

test_that("makeReferenceMarginalDistribution behaves as it should", {
  skip_on_cran()
  skip_on_travis()

  winterThresholded <- winter[winter$O3> median(winter$O3),]
  nMissing <- dim(winter)[1] - dim(winterThresholded)[1]
  
  # gpd fitting
  padded <- as.data.frame(rbind(winterThresholded,
                                cbind(O3=runif(n=nMissing,min=0,max=min(winterThresholded$O3)),
                                      NO2=runif(n=nMissing,min=0,max=min(winterThresholded$NO2)),
                                      NO=runif(n=nMissing,min=0,max=min(winterThresholded$NO)),
                                      SO2=runif(n=nMissing,min=0,max=min(winterThresholded$SO2)),
                                      PM10=runif(n=nMissing,min=0,max=min(winterThresholded$PM10)))))
                                
  
  x.gpd <- migpd(padded,mqu=.7,  penalty="none")
  r.gpd <- migpd(winter,mqu=.7,  penalty="none")
  ref.gpd <- makeReferenceMarginalDistribution(x.gpd,r.gpd,whichNoChange=1)
  
  depEstWithReferenceMargin <- mexDependence(x.gpd,which=1,dqu=0.7,referenceMargin=ref.gpd)
  depEstWithOriginalMargin <-  mexDependence(r.gpd,which=1,dqu=0.7)
  
  expect_equivalent(depEstWithReferenceMargin$dep, depEstWithOriginalMargin$dep)
  
  set.seed(10)
  predictionWithReferenceMargin <- predict(depEstWithReferenceMargin)
  set.seed(10)
  predictionwithOriginalMargins <- predict(depEstWithOriginalMargin)

  expect_equivalent(predictionWithReferenceMargin,predictionWithReferenceMargin)
}
)