R/Sampling_function.R

In A-Morariu/psmplr:

### psmplr package ###
#
# Posterior distribution sampling framework created to take in an INLA model
# and output a sample to mimic all configurations of the model
#
# Main file containing computationally heavy multiplications and sampling calls
#

# parallel multiplications
selectedMeans <- function(lst_of_means, Amat) {
    return(purrr::map(lst_of_means, function(xx) extractEffectMeans(xx, Amat)))
}

selectedCovMat <- function(lst_of_cov, Amat) {
    return(purrr::map(lst_of_cov, function(xx) extractEffectCovMat(xx, Amat)))
}

# sampling
psmplr <- function(inla_model, effect_name, n = 1, constraint_point = 2) {
    # Part 1 - make the selection matrix
    Amat <- makeAMat(inla_model, effect_name, constraint_point)

    # Part 2 - arguments of MVN sampling function
    n <- sampSizes(inla_model, n)
    mu <- extractAllMeans(inla_model) %>% selectedMeans(Amat)
    Sigma <- extractAllCovMat(inla_model) %>% selectedCovMat(Amat)

    # Part 3 - perform sampling (NOTE: NO FORMATTING DONE YET)
    purrr::map2(.x = mu, .y = Sigma, ~ MASS::mvrnorm(n = n, mu = .x, Sigma = .y))
}