projpred (version 2.4.0)

as.matrix.projection: Extract projected parameter draws

Description

This is the as.matrix() method for projection objects (returned by project(), possibly as elements of a list). It extracts the projected parameter draws and returns them as a matrix.

Usage

# S3 method for projection
as.matrix(x, nm_scheme = "auto", ...)

Value

An \(S_{\mathrm{prj}} \times Q\) matrix of projected draws, with \(S_{\mathrm{prj}}\) denoting the number of projected draws and \(Q\) the number of parameters.

Arguments

x

An object of class projection (returned by project(), possibly as elements of a list).

nm_scheme

The naming scheme for the columns of the output matrix. Either "auto", "rstanarm", or "brms", where "auto" chooses "rstanarm" or "brms" based on the class of the reference model fit (and uses "rstanarm" if the reference model fit is of an unknown class).

...

Currently ignored.

Details

In case of the augmented-data projection for a multilevel submodel of a brms::categorical() reference model, the multilevel parameters (and therefore also their names) slightly differ from those in the brms reference model fit (see section "Augmented-data projection" in extend_family()'s documentation).

Examples

if (requireNamespace("rstanarm", quietly = TRUE)) {
  # Data:
  dat_gauss <- data.frame(y = df_gaussian$y, df_gaussian$x)

  # The "stanreg" fit which will be used as the reference model (with small
  # values for `chains` and `iter`, but only for technical reasons in this
  # example; this is not recommended in general):
  fit <- rstanarm::stan_glm(
    y ~ X1 + X2 + X3 + X4 + X5, family = gaussian(), data = dat_gauss,
    QR = TRUE, chains = 2, iter = 500, refresh = 0, seed = 9876
  )

  # Projection onto an arbitrary combination of predictor terms (with a small
  # value for `nclusters`, but only for the sake of speed in this example;
  # this is not recommended in general):
  prj <- project(fit, solution_terms = c("X1", "X3", "X5"), nclusters = 10,
                 seed = 9182)
  prjmat <- as.matrix(prj)
  ### For further post-processing (e.g., via packages `bayesplot` and
  ### `posterior`), we will here ignore the fact that clustering was used
  ### (due to argument `nclusters` above). CAUTION: Ignoring the clustering
  ### is not recommended and only shown here for demonstrative purposes. A
  ### better solution for the clustering case is explained below.
  # If the `bayesplot` package is installed, the output from
  # as.matrix.projection() can be used there. For example:
  if (requireNamespace("bayesplot", quietly = TRUE)) {
    print(bayesplot::mcmc_intervals(prjmat))
  }
  # If the `posterior` package is installed, the output from
  # as.matrix.projection() can be used there. For example:
  if (requireNamespace("posterior", quietly = TRUE)) {
    prjdrws <- posterior::as_draws_matrix(prjmat)
    print(posterior::summarize_draws(
      prjdrws,
      "median", "mad", function(x) quantile(x, probs = c(0.025, 0.975))
    ))
  }
  ### Better solution for post-processing clustered draws (e.g., via
  ### `bayesplot` or `posterior`): Don't ignore the fact that clustering was
  ### used. Instead, resample the clusters according to their weights (e.g.,
  ### via posterior::resample_draws()). However, this requires access to the
  ### cluster weights which is not implemented in `projpred` yet. This
  ### example will be extended as soon as those weights are accessible.
}