prior: Construction of Prior Distributions

In evdbayes: Bayesian Analysis in Extreme Value Theory

Construction of Prior Distributions

Description

Constructing prior distibutions for the location, scale and shape parameters using normal, beta or gamma distributions. A linear trend for the location can also be specified, using a prior normal distribution centered at zero for the trend parameter.

Usage

prior.prob(quant, alpha, trendsd = 0)
prior.quant(prob = 10^-(1:3), shape, scale, trendsd = 0)
prior.norm(mean, cov, trendsd = 0)
prior.loglognorm(mean, cov, trendsd = 0)

Arguments

quant, alpha	Numeric vectors of length three and four respectively. Beta prior distibutions are placed on probability ratios corresponding to the quantiles given in quant.
prob, shape, scale	Numeric vectors of length three. Gamma prior distibutions, with parameters shape and scale, are placed on quantile differences corresponding to the probabilities given in prob.
mean, cov	The prior distibution for the location, log(scale) and shape is taken to be trivariate normal, with mean mean (a numeric vector of length three) and covariance matrix cov (a symmetric positive definite three by three matrix).
trendsd	The standard deviation for the marginal normal prior distribution (with mean zero) placed on the linear trend parameter for the location. If this is zero (the default) a linear trend is not implemented.

Details

See the user's guide.

Value

Returns an object of class "evprior", which is essentially just a list of the arguments passed.

See Also

posterior, pplik

Examples

mat <- diag(c(10000, 10000, 100))
prior.norm(mean = c(0,0,0), cov = mat, trendsd = 10)
prior.quant(shape = c(38.9,7.1,47), scale = c(1.5,6.3,2.6))
prior.prob(quant = c(85,88,95), alpha = c(4,2.5,2.25,0.25))

evdbayes documentation built on March 7, 2023, 5:34 p.m.