lik: Initialize likelihood

In gplite: General Purpose Gaussian Process Modelling

Initialize likelihood

Description

Functions for initializing the likelihood (observation model) which can then be passed to gp_init.

Usage

lik_gaussian(sigma = 0.5, prior_sigma = prior_logunif())

lik_bernoulli(link = "logit")

lik_binomial(link = "logit")

lik_betabinom(link = "logit", phi = 1, prior_phi = prior_logunif())

lik_poisson(link = "log")

Arguments

sigma	Initial value for the noise standard deviation.
prior_sigma	Prior for hyperparameter sigma. See priors.
link	Link function if the likelihood supports non-identity links. See Details for information about possible links for each likelihood.
phi	The over dispersion parameter for beta binomial likelihood.
prior_phi	Prior for hyperparameter phi. See priors.

Details

The supported likelihoods are:

lik_gaussian

Gaussian likelihood. Has no links (uses identity link).

lik_bernoulli

Bernoulli likelihood. Possible links: 'logit' or 'probit'.

lik_binomial

Binomial likelihood. Possible links: 'logit' or 'probit'.

lik_betabinom

Beta binomial likelihood. Possible links: 'logit' or 'probit'.

lik_poisson

Poisson likelihood. Possible links: 'log'.

Value

The likelihood object.

Examples

# Basic usage
cf <- cf_sexp()
lik <- lik_binomial()
gp <- gp_init(cf, lik)

gplite documentation built on Aug. 24, 2022, 9:07 a.m.