# NOT RUN {
## Example 1
x <- c(5, 2, 4, 2, 3, 2, 2, 2, 1, 2) # test data
edf_plot(x, verticals = FALSE) # the 'mathematical' version
edf_plot(x, do.points = FALSE) # good for many sample points
edf_plot(x) # the default
edf_plot(x, log = "x") # logarithmic
(edf_plot(x, log = "x")) # ... with return value
## Note: flat part below first jump cannot be shown in log-scale
## Example 2
data(fire)
u <- 10 # threshold
exceed <- fire[fire > u] # exceedances
excess <- exceed - u # excesses
fit <- fit_GPD_MLE(excess) # fit GPD to excesses
## Plot empirical distribution function of excesses
## (partially with fitted GPD overlaid)
edf_plot(excess)
res <- edf_plot(excess, log = "x")
z <- tail(res$t, n = -1)
lines(z, pGPD(z, shape = fit$par[["shape"]], scale = fit$par[["scale"]]))
## Plot empirical distribution function of exceedances
## (partially with fitted GPD overlaid)
edf_plot(exceed)
res <- edf_plot(exceed, log = "x")
z <- tail(res$t, n = -1)
lines(z, pGPD(z-u, shape = fit$par[["shape"]], scale = fit$par[["scale"]]))
## Note: Q-Q plots are typically more meaningful
qf <- function(p) # quantile function of df
qGPD(p, shape = fit$par[["shape"]], scale = fit$par[["scale"]])
qq_plot(excess, FUN = qf)
qq_plot(exceed, FUN = function(p) u + qf(p))
# }
Run the code above in your browser using DataLab