sample_with_constraints: Sample with constraints

In KatarzynaReluga/postcAIC: Post-cAIC selection inference under Linear Mixed Models

Description

Function sample_with_constraints provides samples from a constrained normal distribution. Constraints are imposed due to the data driven model selection.

Usage

sample_with_constraints(
  n_starting_points,
  p,
  n,
  n_models_to_compare,
  list_constraints,
  n_samples = 10000,
  burn.in = 1000,
  scale_mvrnorm
)

Arguments

n_starting_points	Number of initial starting points for sampling from a truncated normal distribution
p	Number of all fixed parameters under consideration (intercept included)
n	Number of clusters (random effects)
n_models_to_compare	Number of models in a model set to compare with (all models in the model set minus the selected model)
list_constraints	List which describes the constraints imposed on the normal distribution
n_samples	Number of samples from a constrained distribution
burn.in	The number of burn-in iterations. The Markov chain is sampled n_samples + burn.in times, and the last n samples are returned.
scale_mvrnorm	Scale parameter for multivariate normal distribution to sample. Default: scale_mvrnorm = 10.

Details

• n_starting_points - in practice we use 2 times n_starting_points, because we sample using initial values and negative initial values obtained applying function find_starting_points

• burn.in - parameter of function rtmg

Value

List with elements:

sample_fixed_full	n_samples from a multivariate constrained normal distribution to obtain critical values to construct confidence intervals for fixed effects
sample_random	n_samples from a multivariate normal distribution to obtain critical values to construct confidence intervals for random effects
sample_mix_full	n_samples of sample_fixed_full and sample_random stacked together

KatarzynaReluga/postcAIC documentation built on Jan. 25, 2022, 12:33 a.m.