rstan_default_init: Helper Function to Create the Default 'rstan' scalar init

In maomlab/BayesPharma: Tools for Bayesian Analysis of Non-Linear Pharmacology Models

Helper Function to Create the Default rstan scalar init

Description

By default, rstan will initialize parameters uniformly at random in the range (-2, 2), on the unconstrained scale. Description of how stan transforms parameters to satisfy constraints is described in the stan documentation

This helper is especially useful for running models using the cmdstanr backend, which requires all parameters (including distributional) parameters to be initialized.

Usage

rstan_default_init(lb = NULL, ub = NULL, dim = 0)

Arguments

lb	numeric lower bound for parameter
ub	numeric upper bound for parameter
dim	numeric dimension of parameter.

Value

function that return the default brms initial value for a parameter. If dim=0, then it will be a numeric scalar, if dim=1 or greater, than return an array with the given dimension.

See Also

rstan::stan

Examples

## Not run: 

# Explicitly set the default initialization for the distributional parameter
# 'sigma' when family=gaussian().
init <- BayesPharma::sigmoid_antagonist_init(
  sigma = BayesPharma::rstan_default_scalar_init(lb = 0))


## End(Not run)

maomlab/BayesPharma documentation built on Aug. 24, 2024, 8:45 a.m.