How the hierarchy changes across target effect-modifier means

The component effects are beta + Gamma x, so a component's rank is a function of the target means x and components can cross. A single hierarchy quoted without its target means is therefore incomplete. This plot shows the family of average conditional link-scale hierarchies across a mean grid. It does not standardize over different population distributions.

Usage

plot_rank_curve(
  x,
  em = NULL,
  values = NULL,
  at = NULL,
  what = c("treatment", "component"),
  lower_is_better = FALSE,
  metric = c("sucra", "mean_rank", "p_best"),
  ...
)

Arguments

x: A cmlnmr() fit, or the data frame returned by rank_curve().
em: Name of the effect modifier to vary. Required when x is a fit.
values: Numeric vector of target values for em. Required when x is a fit.
at: Optional named vector fixing the other effect modifiers.
what, lower_is_better: See cpaic_ranks().
metric: Which ranking metric to trace: "sucra" (default), "mean_rank", or "p_best".
...: Passed to rank_curve() when x is a fit.

Value

A ggplot object.

Details

There is no counterpart in multinma, which ranks in one population at a time.

References

Wigle A, Beliveau A, Nikolakopoulou A, Lin L (2026). Creating Treatment and Component Hierarchies in Component Network Meta-Analysis.

Examples

if (FALSE) {
plot_rank_curve(fit, em = "x1", values = seq(-1, 1, by = 0.25),
                what = "component")
}