Plot how much each edge informs a chosen contrast

Visualizes edge_influence(): the weight with which each observed edge enters the estimate of one relative effect. This plot exists because the usual population-adjustment diagnostics cannot see the failure it detects. An IPD edge with zero influence on your contrast contributes nothing to it, so reweighting that edge cannot move the answer, however healthy its effective sample size looks. Such edges are drawn in red and labelled.

Usage

plot_edge_influence(object, treatment, comparator = NULL, ...)

Arguments

object

A cpaic_bridge, cpaic_maic, or cpaic_stc object.

treatment, comparator

The contrast of interest. comparator defaults to the network reference.

...

Passed to edge_influence() (for example tol).

Value

A ggplot object.

Details

There is no counterpart to this plot in multinma; it is specific to the component bridge, where a contrast is a weighted combination of edges chosen by the component design rather than by the network path.

See also

edge_influence(), effective_sample_size()

Examples

net <- cpaic_network(cpaic_bin_agd, ipd = cpaic_bin_ipd, sm = "OR",
                     family = "binomial", ipd_covariates = "x1",
                     inactive = "Placebo")
br <- cnma_bridge(net)
plot_edge_influence(br, treatment = "A+B+C")