Skip to contents

Posterior prediction draws for mcpfit objects

Source: R/mcpfit_methods.R
posterior_epred.mcpfit.Rd

Methods for the {rstantools} posterior-prediction generics. They return a draws-by-observation matrix and enable {tidybayes} workflows such as add_epred_draws(), add_predicted_draws(), and add_linpred_draws(). These methods and workflows require the suggested package {rstantools}.

Usage

posterior_epred.mcpfit(
  object,
  newdata = NULL,
  draws = NULL,
  ndraws = NULL,
  re.form = NULL,
  re_formula = NULL,
  dpar = NULL,
  seed = NULL,
  ...
)

posterior_predict.mcpfit(
  object,
  newdata = NULL,
  draws = NULL,
  ndraws = NULL,
  re.form = NULL,
  re_formula = NULL,
  seed = NULL,
  ...
)

posterior_linpred.mcpfit(
  object,
  transform = FALSE,
  newdata = NULL,
  draws = NULL,
  ndraws = NULL,
  re.form = NULL,
  re_formula = NULL,
  dpar = NULL,
  seed = NULL,
  ...
)

Arguments

object

An mcpfit object.

newdata

Optional data frame at which to evaluate the model. For GARMA posterior_predict(), only predictors and required response auxiliaries are needed: each response series is generated recursively without conditioning on an observed response column.

draws, ndraws

Number of posterior draws to return. draws follows the {rstantools} convention; ndraws is the mcp spelling. Supply at most one.

re.form, re_formula

Group-level effects to include. NULL includes all effects and NA excludes them.

dpar

Distributional parameter for posterior_epred() and posterior_linpred(); NULL uses the expected response.

seed

Optional integer seed for draw selection and posterior prediction.

...

Must be empty. Reserved for future use.

transform

For posterior_linpred(), return the inverse-link transformed expected response instead of the linear predictor.

Value

A numeric N_draws by nrow(newdata) matrix.

Details

For GARMA models, posterior_predict() generates each replicated response series recursively. It does not condition later predictions on the observed response history, unlike fitted() and posterior predict(). These methods require posterior draws. For prior prediction, use predict() with prior = TRUE, which also generates fresh series.

For binomial models, posterior_epred() and posterior_predict() (and corresponding {tidybayes} workflows such as add_epred_draws()) follow {brms} and {rstantools} conventions by returning values on the outcome count scale (rate = FALSE), i.e., expected counts \(E[Y] = n\mu\) and simulated counts in \(\{0, \dots, n\}\). In contrast, fitted() and predict() default to proportions (rate = TRUE). To obtain the success probability parameter \(\mu\) on the \([0, 1]\) scale regardless of trial counts, pass dpar = "mu".

See also

fitted.mcpfit(), predict.mcpfit()