mpp_derivatives: Multivariate Probit Event Probability Derivatives

In finnlindgren/multiprobit: Bayesian Multivariate Probit Regression

Multivariate Probit Event Probability Derivatives

Description

FUNCTION_DESCRIPTION

Usage

mpp_gradient_mu(
  y,
  mu,
  ...,
  gaussint_options = NULL,
  h = 1e-06,
  symmetric = FALSE
)

mpp_hessian_mu(
  y,
  mu,
  ...,
  gaussint_options = NULL,
  h = 1e-04,
  diagonal = FALSE
)

mpp_gradient_u(
  y,
  mu,
  u,
  Sigma_model,
  gaussint_options = NULL,
  h = 1e-06,
  symmetric = FALSE,
  log = FALSE,
  ...
)

Arguments

y	A matrix of multivariate 0/1 observations
mu	A matrix of matrix of multivariate latent scale expectation parameters
...	Further parameters passed on to mpp()
gaussint_options	list of options for excursions::gaussint. By default sets seed = 1L to ensure consistent approximation error.
h	Step size for finite differences, Default: 1e-06 (for gradients) or 1e-04 (for hessian)
symmetric	For gradients, whether to use symmetric finite differences, Default: FALSE
diagonal	Logical; if TRUE, only the diagonal of the hessian is evaluated, Default: FALSE
u	A vector of latent variables identifying the Normalised Wishart matrix, length d(d-1)/2
Sigma_model	A wm_model object
log	Whether to compute gradient of the log-probability, Default: FALSE

Details

DETAILS gradient

DETAILS hessain

DETAILS

Value

OUTPUT_DESCRIPTION gradient for mu

OUTPUT_DESCRIPTION hessian for mu

OUTPUT_DESCRIPTION gradient for u

See Also

mpp()

Examples

## Not run: 
if (interactive()) {
  # EXAMPLE1 gradient
}

## End(Not run)
## Not run: 
if (interactive()) {
  # EXAMPLE1 hessian
}

## End(Not run)
## Not run: 
if (interactive()) {
  # EXAMPLE1
}

## End(Not run)

finnlindgren/multiprobit documentation built on June 14, 2025, 1:12 p.m.