Predict Method for rhierNegbinRw Objects

Computes coefficient-level or predictive count-response draws.

Usage

# S3 method for class 'rhierNegbinRw'
predict(
  object,
  newdata = NULL,
  type = "DeltaZ+mu",
  burn = 0,
  mode = "coefficients",
  nsim = 10,
  r_verbose = TRUE,
  force_tree_eval = FALSE,
  ...
)

Arguments

object

A fitted rhierNegbinRw object.

newdata

A list whose required fields depend on mode:

• For mode = "coefficients": requires newdata$Z (if the model was fit with Z).

• For mode = "posterior": requires newdata$regdata, a list with per-unit design matrices X matching the fitted units.

• For mode = "prior": requires newdata$X, a list of design matrices for prediction units, and newdata$Z when the model was fit with Z.

type

Character; interpretation depends on mode:

• For mode = "coefficients": one of "DeltaZ", "DeltaZ+mu", "SigmaZ".

• For mode %in% c("prior","posterior"): must be "response".

burn

Integer, number of initial MCMC draws to discard.

mode

Prediction mode: "coefficients" (default), "posterior", or "prior".

nsim

Integer, number of Monte Carlo draws for prior predictive response mixing.

r_verbose

Logical, print progress updates?

...

Additional arguments passed to pwbart for BART models.

Value

Depends on mode / type:

• For mode = "coefficients", type %in% c("DeltaZ", "DeltaZ+mu"): 3D array [npred, ncoef, ndraws_out] of predicted betabar values.

• For mode = "coefficients", type = "SigmaZ": 4D array [npred, ncoef, ncoef, ndraws_out] of covariance draws at each prediction unit.

• For mode %in% c("prior","posterior"), type = "response": list of matrices with simulated count draws per unit (nobs_i x ndraws_out).