parallel_greedy_search: Routine to perform greedy search at the beginning of MASSIVE...

In igbucur/MASSIVE: MASSIVE: Tractable and Robust Bayesian Learning of Many-Dimensional Instrumental Variable Models

Description

At each greedy step, we examine all of the IV models in the neighbor set in parallel. The neighbor with the highest model evidence is chosen and the algorithm moves to it if the evidence is higher compared to the current IV model.

Usage

parallel_greedy_search(
  J,
  N,
  SS,
  sigma_G,
  sd_slab = 1,
  sd_spike = 0.01,
  init_model = NULL,
  get_neighbors = neighbor_IV_models,
  LA_function = safe_Laplace_approximation,
  ...
)

Arguments

J	Integer number of candidate instruments..
N	Integer number of observations.
SS	Numeric matrix containing first- and second-order statistics.
sigma_G	Numeric vector of genetic IV standard deviations.
sd_slab	Numeric scale parameter of slab component.
sd_spike	Numeric scale parameter of spike component.
init_model	Character vector describing starting IV model in search.
get_neighbors	Function to get neighbor IV models at every step.
LA_function	Function for computing the IV model Laplace approximation
...	Extra arguments to pass to Laplace approximation function.

Value

A list containing the optimum found with greedy search, the list of IV models visited and their approximations, and the number of visited models.

Examples

J <- 5 # number of instruments
N <- 1000 # number of samples
parameters <- random_Gaussian_parameters(J) 
EAF <- runif(J, 0.1, 0.9) # EAF random values
dat <- generate_data_MASSIVE_model(N, 2, EAF, parameters)
parallel_greedy_search(J, N, dat$SS, binomial_sigma_G(dat$SS), 1, 0.01)

igbucur/MASSIVE documentation built on Oct. 26, 2020, 1:26 a.m.