Pseudo-confidence-interval contours for a funnel plot

geom_funnel_contour() draws the funnel-shaped pseudo confidence interval(s) around a reference effect. For a study with standard error se, the expected effect under no bias lies within centre +/- z * se, so the contours are straight lines fanning out from the apex at se = 0. On a funnel plot with a reversed standard-error axis they form the familiar funnel.

Usage

geom_funnel_contour(
  centre,
  se_max,
  level = 0.95,
  ...,
  colour = "grey55",
  linetype = "dashed",
  linewidth = 0.4,
  alpha = NA
)

Arguments

centre: Reference effect (usually the pooled estimate) the funnel is centred on, on the analysis scale.
se_max: Largest standard error the contours should reach (the bottom of the funnel).
level: Confidence level(s) for the contour(s), e.g. 0.95 or c(0.95, 0.99). Default 0.95.
...: Other arguments passed on to ggplot2::layer().
colour, linetype, linewidth, alpha: Line appearance.

Value

A ggplot2 layer.

Details

The layer builds its own data from centre, se_max, and level, so it does not use the plot's data; add it to a plot that maps the effect to x and the standard error to y.

Examples

library(ggplot2)
studies <- data.frame(
  estimate = c(-0.5, -0.2, -0.4, 0.1, -0.3),
  se       = c(0.10, 0.30, 0.15, 0.35, 0.22)
)
ggplot(studies, aes(x = estimate, y = se)) +
  geom_funnel_contour(centre = -0.3, se_max = 0.4) +
  geom_point() +
  scale_y_reverse()