gpd_fit: Fit a time serie with a GPD distribution .
In thaos/FARg: FAR estimation

Description Usage Arguments Details Value Examples

gpd_fit fits a time serie with a GPD distribution a where scale parameter depend linearly on a set of covariates.

1 2	gpd_fit(y, data, mu_mod = ~1, sig_mod = ~1, sig_link = "log", time_var, qthreshold, init = NULL)

`y`	the time serie to be fitted.
`data`	a data.frame object with where the function looks first for the variables y, time_var and the covariates specified in the mu_mod and sig_mod arguments.
`mu_mod`	a formula defining the covariates to be used in quantile regression to set the threshold of the GPD.
`sig_mod`	a formula defining the covariates the scale parameter of the GPD depends linearly on.
`sig_link`	a link function name for the parameter sigma: sig_link(sigma) is a linear function of the covariates.
`time_var`	a variable used to define the time in the time serie. It can also be a string giving the variable name.
`qthreshold`	the level of quantile used to set the GPD threshold.
`init`	vector of initialization parameter for the minimization of the negative log-likelihood. if NULL, the initialisation is done using the function fevd from the extRemes packages.

MLE fit of a time serie y using a GPD distribution where the scale parameter depen linearly on a set of covariates. The threshold is defined using quantile regression (function rq from the quantreg package. The optimization of the negative log-likelihood is done with nlminb function in R.

returns an object of class gpd_fit. It contains the nlminb output which provides the estimated parameters as well the minimum of the negative log-likelihood. The arguments use to call gpd_fit are included in the list as well.

data(tas)
#Example with the same covariate for the mean and variance parameter
gp_fit <- gpd_fit(eur_tas, data=tas, mu_mod=~gbl_tas, sig_mod=~gbl_tas, time_var="year", qthreshold=0.9)
# get the values of the mean and variance parameters of the GPD at each time
compute_par(gp_fit, tas)
# plot diagnostic plot of the fit : qqplot, density of fitted vs theorical density, times series ans return levels
plot(gp_fit)