Forest plot of relative or component effects

A ggplot2 forest plot of the estimates in a cpaic fit. Ported from multinma::plot.nma_summary() (Phillippo et al. 2020) and re-implemented on ggplot2 alone.

Usage

forest(
  x,
  ...,
  what = c("relative", "component"),
  order = c("estimate", "alphabetical", "none"),
  ref_line = NULL,
  point_size = 1.2,
  show_na = TRUE
)

# S3 method for class 'cpaic_effects'
plot(x, y, ...)

# S3 method for class 'cpaic_bridge'
plot(x, y, ...)

# S3 method for class 'cpaic_fit'
plot(x, y, ...)

Arguments

x: A cpaic_effects data frame (from relative_effects()), a fitted cpaic object (cpaic_bridge, cpaic_maic, cpaic_stc, cpaic_mlnmr), or a component-effect data frame from component_effects().
...: Passed to relative_effects() / component_effects() when x is a fit (for example newdata for a cmlnmr() fit).
what: "relative" (default) for relative effects, or "component" for the incremental effect of each component.
order: Row ordering: "estimate" (default, most to least favorable), "alphabetical", or "none" (the order in the input).
ref_line: Position of the vertical reference line. Defaults to the null value of the summary measure (1 on a back-transformed ratio scale, 0 otherwise); NA draws none.
point_size: Size of the point-estimate marker. Default 1.2.
show_na: Show non-estimable contrasts as labelled empty rows? Default TRUE. Setting this to FALSE hides evidence that the network cannot answer part of your question, so leave it on unless you have a reason.
y: Unused, for compatibility with the plot() generic.

Value

A ggplot object.

Details

Contrasts that the component design cannot identify are shown, labelled not estimable, rather than silently dropped. Dropping them would leave the reader with a plot that looks complete when it is not; see estimable_effects() and estimable_effects_at().

A table with several comparators (from relative_effects(all_contrasts = TRUE)) is faceted, one panel per comparator. Ratio measures are drawn on a log axis, on which they are symmetric. Pass level through ... to change the interval width.

Examples

net <- cpaic_network(cpaic_bin_agd, sm = "OR", inactive = "Placebo")
br <- cnma_bridge(net)
forest(br)

forest(br, what = "component")