Wrapper for estimating correlation matrix (multivariate normal distribution)

Estimates a correlation matrix (along with means and standard deviations) assuming a multivariate normal distribution from observation data. If there are 2 observed variables, it automatically switches to estimate the scalar correlation coefficient (`corr`) directly.

Usage

rtmb_corr(
  data,
  prior = list(lkj_eta = 1, mu_sd = 10, sigma_rate = 1),
  init = NULL,
  null = NULL
)

Arguments

data: Observation data frame or matrix (N x P).
prior: List of hyperparameters for prior distributions. Default is list(lkj_eta = 1.0, mu_sd = 10, sigma_rate = 1.0).
init: List of initial values (optional).
null: Target when creating a null model (e.g., "corr"). Optional.

Value

An instance of the RTMB_Model class.

Examples

if (FALSE) { # \dontrun{
  # Simulate bivariate normal data with a true correlation of 0.5
  set.seed(123)
  N <- 50
  rho <- 0.5
  cov_mat <- matrix(c(1, rho, rho, 1), nrow = 2)

  if (requireNamespace("MASS", quietly = TRUE)) {
    data_corr <- MASS::mvrnorm(N, mu = c(0, 0), Sigma = cov_mat)
    colnames(data_corr) <- c("X1", "X2")

    fit_corr <- rtmb_corr(data = data_corr)

    mcmc_corr <- fit_corr$sample(sampling = 500, warmup = 500, chains = 2)
    mcmc_corr$summary()

    bf_corr <- mcmc_corr$bayes_factor(null_model = "corr")
    print(bf_corr)
  }
} # }