Skip to contents

CAMS subgroup and overall effects on an information-fraction grid

Source: R/contribution.R
predict_cams_by_if.Rd

Evaluates the fitted CAMS posterior at a grid of information fractions without refitting. Each posterior row already stores the joint pairing of \((\alpha,\theta)\) (overall block) with the independent interaction draw \(\gamma\). For draw \(s\) and grid value \(\pi_k\): $$M^{(s)}(\pi_k)=\alpha^{(s)}+\theta^{(s)}\pi_k,$$ $$\mu_A^{(s)}(\pi_k)=M^{(s)}(\pi_k)-\pi_k\gamma^{(s)},$$ $$\mu_B^{(s)}(\pi_k)=M^{(s)}(\pi_k)+(1-\pi_k)\gamma^{(s)}.$$ Stored b_Intercept is \(\alpha\), stored b_ifrac is \(\beta=\theta-\gamma\), and stored b_subgroup* is \(\gamma\), so \(\theta=\beta+\gamma\). Subgroup lines are parallel with vertical gap \(\gamma\).

Usage

predict_cams_by_if(
  fit,
  if_grid = NULL,
  level = 0.95,
  ifrac_var = NULL,
  resolution = 200,
  max_support = NULL,
  exponentiate = FALSE,
  interval.type = NULL
)

Arguments

fit

A interactmeta from cams().

if_grid

Numeric information-fraction grid. If NULL, spans the observed study IF range with a small margin (see plot_contribution()).

level

Coverage probability for lower / upper (default 0.95).

ifrac_var

ifrac_* column used to build the default grid.

resolution

Grid length when if_grid is NULL.

max_support

Optional cap on unweighted posterior rows. Rejected for Bayesian CAMS DIRECT supports; omit it.

exponentiate

If TRUE, summarize on the exponential scale.

interval.type

"central" or "shortest" mixture interval on Bayesian DIRECT nodes (default: fit$interval.type, usually "central"). Pass "shortest" for the narrowest equal-probability interval.

Value

A data frame with one row per grid point and subgroup: ifrac, subgroup, estimate, lower, upper, plus overall_estimate, overall_lower, overall_upper for \(M(\pi)=\alpha+\theta\pi\) at that grid point.

Details

Pointwise estimate / lower / upper use mixture intervals from the DIRECT-node Gaussian mixture (weights .weight, within-node SDs .cond_*). Alpha and theta stay on the joint overall-block support (never convolved with each other).

See also

modifiers(), combined(), plot_contribution()