ppreg: Point process generalized linear regression modelling
In pengyuwei94/evreg: Extreme value regression modelling

Description Usage Arguments Details Value References

Regression modelling with a Point Process response distribution and generalized linear modelling of each parameter, specified by a symbolic description of each linear predictor and the inverse link function to be applied to each linear predictor. Quantile regression has been used to set a threshold for which the probability p of threshold exceedance is approximatly constant across different values of covarites.

1
2
3

ppreg(y, data, p = 0.5, npy = 365, mu = ~1, sigma = ~1, xi = ~1,
  mustart, sigmastart, xistart, invmulink = identity,
  invsigmalink = identity, invxilink = identity, ...)

`y`	Either a numeric vector or the name of a variable in `data`. `y` must not have any missing values.
`data`	A data frame containing `y` and any covariates. Neither `y` nor the covariates may have any missing values.
`p`	Probability quantile. This is generally a number strictly between 0 and 1.
`mu, sigma, xi`	Formulae (see `formula`) for the PP parameters `mu` (location), `sigma` (scale), e.g. `mu = ~ x`. Parameter `xi` (shape) is holded as fixed.

Fitting a point process regression model allows us to relex the stationary assumption. The function fits a point process regression model by setting a covariate-dependent threshold using a linear quantile regression. The goal is to keep the probability p of threshold exceedances retains constant across different values of the covariates. The form of the covariate-dependent threshold is that of Northrop and Jonathan (2011) Eq (5).

Warning. At the moment, only models with identity inverse link function for all parameters and constant shape parameter can be fitted. Different optimization methods may result in wildly different parameter estimates.

An object (list) of class c("gev", "evreg"), which has the following components

`coefficients`	A named numeric vector of the estimates of the model parameters.
`se`	estimated standard errors

Northrop, Paul J, and Philip Jonathan. 2011. "Threshold Modelling of Spatially Dependent Non-Stationary Extremes with Application to Hurricane-Induced WaveHeights." Environmetrics22 (7). Wiley Online Library: 799-809.

pengyuwei94/evreg documentation built on Aug. 29, 2019, 1:06 p.m.