add_priors: Add Hyperparameter Values to Standata Object

In jpkeller/bercs: Bayesian Exposure-Response Curves via Stan

Description

Sets prior distribution hyperparameters using defaults or provided values

Usage

add_priors(standata, ...)

add_priors_exposure_model(
  standata,
  etaG = 0,
  sigmaG = c(0, 1),
  reK = 0,
  sigmaK = c(0, 1),
  reI = 0,
  sigmaI = c(0, 1),
  sigmaW = c(0, 1),
  theta = 0,
  sigmaTheta = c(0, 1),
  ...
)

add_priors_outcome_model(
  standata,
  beta = 0,
  sigmaBeta = c(0, 1),
  gamma = 0,
  sigmaGamma = c(0, 1),
  delta = 0,
  sigmaDelta = c(0, 1),
  sigmaI = c(0, 1),
  beta_nu = 1,
  ...
)

Arguments

standata	Either a standata_exposure or standata_outcome object.
...	Arguments passed to add_priors_exposure_model or add_priors_outcome_model. Unlisted arguments are ignored.
etaG	Prior mean and standard deviation for the group means parameter 'etaG'.
sigmaG	Prior mean and standard deviation for the group standard deviation parameter 'sigmaG'.
reK	Prior mean for cluster-level random effects.
sigmaK	Prior mean and standard deviation for the standard deviation of the cluster random effects, 'sigmaK'.
reI	Prior mean for household-level random effects.
sigmaI	Prior mean and standard deviation for the standard deviation of the subject-level random effect.
sigmaW	Prior mean and standard deviation for the residual standard deviation parameter 'sigmaW'.
theta	Prior mean and standard deviation for the time coefficients 'thetaG'.
sigmaTheta	Prior mean and standard deviation for the standard deviation parameter for time trends, 'sigmaTheta'.
beta	Prior mean for the exposure coefficient(s).
sigmaBeta	Prior mean and standard deviation for the standard deviation of exposure coefficient parameter.
gamma	Prior mean for the covariate coefficient(s).
sigmaGamma	Prior mean and standard deviation for the standard deviation of covariate coefficient parameter.
delta	Prior mean for the time coefficient(s).
sigmaDelta	Prior mean and standard deviation for the standard deviation of time coefficient parameter.
beta_nu	Prior value for the LKJ prior on the correlation between exposure coefficients from different studies.

Details

Based on its class, add_priors calls add_priors_exposure_model or add_priors_exposure_model. For the variance parameters, the prior distributions are half-normal distributions parameterized by (mean, sd).

See Also

create_standata_exposure, create_standata_outcome, sample_exposure_model, sample_outcome_model

Examples

# Create simulated data
exp_data <- create_standata_exposure(group=rep(1, 10),
                                     conc=rnorm(10),
                                     unit_id=rep(0:1, 5),
                                     time=runif(10))
# Add comnbination of default and custom prior
exp_data <- add_priors(exp_data,
                       sigmaI=c(0, 0.1),
                       sigmaK=c(0, 2))

jpkeller/bercs documentation built on March 24, 2021, 5:36 a.m.