Skip to contents

Display independently allocated direct and indirect evidence estimates with separate pointwise confidence intervals and numerical columns.

Usage

network_nodesplit_plot(
  object,
  method = c("conditional", "ML"),
  comparisons = NULL,
  level = 0.95,
  adjust = c("none", "holm", "bonferroni"),
  strata = NULL,
  main = "Direct and indirect node-split effects",
  estimable_only = FALSE,
  point_cex = 1,
  label_cex = 0.8
)

Arguments

object

A fitted joint network model.

method

conditional holds fitted heterogeneity fixed; ML refits heterogeneity jointly in each split model. Both use normal Wald inference.

comparisons

Optional data frame with treatment_1 and treatment_2 columns selecting unordered pairs. Selection includes all fitted mean strata; reversing a requested pair does not reverse the returned contrast direction.

level

Coverage of pointwise normal-approximation intervals for direct effects, indirect effects and their difference.

adjust

none, holm, or bonferroni for the requested node-split pair/stratum family, retaining nonestimable hypotheses in its size. Output adds p_adjusted and an adjustment attribute. Intervals are pointwise.

strata

Named list selecting fitted outcome/time strata for display. Adjustment uses all requested comparisons before this display selection.

main

Plot title.

estimable_only

Logical; omit unavailable splits from the figure (default FALSE). Returned rows and the adjustment family are unchanged.

point_cex

Scaling for point markers.

label_cex

Scaling for treatment/axis labels. Font family and style inherit base graphics settings such as par(family="sans", font=1).

Value

Invisibly returns the selected node-split table, including confidence limits, adjusted p-values, estimability and failure reasons.

Details

Calls network_nodesplit() and plots treatment_2 minus treatment_1 for each pair/stratum. The direct and indirect series use separate symbols and vertical offsets on a shared effect axis. Unavailable evidence is labelled not estimable; an available direct estimate is still shown when an indirect route is absent. Intervals are model-based normal intervals, conditional on the fitted covariance or the jointly refitted ML covariance. adjust changes the difference-test p-values in the returned table, not these pointwise intervals. Interval overlap is not the direct-minus-indirect Wald test: use the returned difference interval and p-value, which retain direct-indirect covariance.

These whole-study node splits differ from the descriptive conditional GLS summaries displayed by network_forest_overlay_plot(). For a plot of the direct-minus-indirect difference, use network_incoherence_plot() with display="nodesplit".

Examples

d <- data.frame(s=1:9,a=rep(c("A","A","B"),3),
  b=rep(c("B","C","C"),3),g=rep(c(.2,.6,.2),3),v=.04)
fit <- network_meta(d,"s","a","b","g","v")
network_nodesplit_plot(fit,adjust="holm")