EMC2 (version 2.0.2)

get_prior_factor: Prior specification and prior sampling for factor estimation

Description

To get the default priors for a given design: get_prior_factor(design = design, sample = FALSE)

Usage

get_prior_factor(
  prior = NULL,
  n_pars = NULL,
  sample = TRUE,
  N = 1e+05,
  selection = "mu",
  design = NULL,
  Lambda_mat = NULL,
  n_factors = NULL
)

Value

A list with a single entry of type of samples from the prior (if sample = TRUE) or else a prior object

Arguments

prior

A named list that can contain the prior mean (theta_mu_mean) and variance (theta_mu_var) on the group-level mean; the variance of the loadings (theta_lambda_var); shape and rate of the factor variances (ap and bp) and shape and rate of the residual variances (as and bs). For NULL entries, the default prior is is used.

n_pars

Often inferred from the design, but if design = NULL, n_pars will be used to determine the size of prior.

sample

Whether to sample from the prior or to simply return the prior. Default is TRUE,

N

How many samples to draw from the prior, the default is 1e5

selection

Character. If sample = TRUE, what priors to sample from.

design

The design obtained from design(), required when map = TRUE

Lambda_mat

The loadings constraint matrix.

n_factors

Integer. The number of factors.

Details

For details see Ghosh, J., & Dunson, D. B. (2009). Default prior distributions and efficient posterior computation in Bayesian factor analysis. Journal of Computational and Graphical Statistics, 18, 306-320. or Stevenson, N., Innes, R. J., Gronau, Q. F., Miletic, S., Heathcote, A., PhD, Forstmann, B., & Brown, S. (2024). Using group level factor models to resolve high dimensionality in model-based sampling. https://doi.org/10.31234/osf.io/pn3wv.

Note that if sample = FALSE, prior$theta_mu_invar (the inverse of the prior covariance matrix on the group-level mean) is returned, which is only used for computational efficiency

Examples

# First define a design for the model
design_DDMaE <- design(data = forstmann,model=DDM,
                           formula =list(v~0+S,a~E, t0~1, s~1, Z~1, sv~1, SZ~1),
                           constants=c(s=log(1)))
# Now get the default prior
prior <- get_prior_factor(design = design_DDMaE, sample = FALSE, n_factors = 3)
# We can change values in the default prior or use `prior`
# Then we can get samples from this prior e.g.
samples <- get_prior_factor(prior = prior, design = design_DDMaE,
  sample = TRUE, selection = "mu", n_factors = 3)