spike_and_slab_logistic: Group spike and slab variable selection with Gaussian outcome

In IQSS/moretrees: Multi-Outcome Regression with Tree-Structured Shrinkage

Description

Here's a brief description. spike_and_slab_logistic performs group variable selection via a spike and slab prior for binary data. The posterior is approximated via variational inference. This function returns the parameters of the variational approximation.

Usage

spike_and_slab_logistic(
  dsgn,
  initial_values,
  tol,
  max_iter,
  update_hyper,
  update_hyper_freq,
  hyper_fixed,
  print_freq,
  hyper_random_init,
  vi_random_init
)

Arguments

tol	Convergence tolerance for ELBO.
update_hyper	Update hyperparameters? Default = TRUE.
update_hyper_freq	How frequently to update hyperparameters. Default = every 10 iterations.
y	Integer vector of length n containing outcomes; 1 = success, 0 = failure.
X	matrix of dimension n x sum(K), where n is the number of units, and K[g] is the number of variables in group g.
groups	A list of length G (number of groups), where groups[[g]] is an integer vector specifying the columns of X that belong to group g.
W	matrix of non-sparse regression coefficients of dimension n x m
maxiter	Maximum number of iterations of the VI algorithm.

Details

All the details go here!

Value

A list of variational parameters.

Model Description

Describe group spike and slab prior and all parameters here.

See Also

Other spike and slab functions: compute_betas(), compute_thetas(), ml_by_group(), spike_and_slab_logistic_moretrees(), spike_and_slab_normal_moretrees(), spike_and_slab_normal(), spike_and_slab()

IQSS/moretrees documentation built on March 20, 2020, 8:44 p.m.