Skip to contents

Draw an interactive funnel plot for a meta-analysis

Source: R/meta_funnel.R
ggfunnel.Rd

Draws each study's effect against its standard error, with the most precise studies at the top, to show small-study effects. The shaded contours mark where a study would be statistically significant against no effect (Peters et al. 2008), so a gap in the unshaded area, where studies would not be significant, points to publication bias rather than heterogeneity alone. The solid line is the pooled estimate and the dashed lines around it the region where 95% of studies would fall without heterogeneity or bias.

Usage

ggfunnel(
  x,
  data = NULL,
  rob = NULL,
  hover = NULL,
  contours = c(0.1, 0.05, 0.01),
  trim_fill = FALSE,
  tests = TRUE,
  exponentiate = NULL,
  xlim = NULL,
  xlab = NULL,
  title = NULL,
  caption = NULL,
  family = "Lato"
)

Arguments

x

A fitted meta-analysis: an rma.uni object from metafor::rma(), or a meta object from the 'meta' package, such as the result of meta::metabin() or meta::metagen(). Models with moderators are not supported.

data

Optional data frame with one row per study, in the order of the model, holding the columns to show. Defaults to the data stored in the fit, which is there when the model was fitted with a data argument.

rob

Name of the column of data holding each study's overall risk of bias judgment, which colors its point. Judgments are matched by their wording, as in ggmeta().

hover

Names of columns of data shown in each study's hover card. Defaults to columns.

contours

Significance levels for the shaded contours, against no effect. NULL draws none.

trim_fill

Add the studies imputed by trim and fill and the adjusted estimate.

tests

Add the tests for small-study effects, in a collapsed section under the plot.

exponentiate

Show effects on the ratio scale. Defaults to TRUE for ratio measures (RR, OR, HR, IRR, ROM and Peto odds ratios), which are modeled on the log scale.

xlim

Optional limits of the effect axis, on the scale shown.

xlab

Axis label. Defaults to the name of the effect measure.

title, caption

Title above the plot and note below it.

family

Font family. The package ships Lato and registers it on load.

Value

An object of class ggfunnel, which prints as an interactive widget. Use graph_widget(), graph_plot() or graph_save() for the widget, a static ggplot or a file. The field tests holds the tests as a data frame.

Details

Hovering over a study shows its effect, weight, risk of bias, the columns named in hover and the significance zone it falls in. Clicking it opens the pooled estimate with that study left out, beside its full record.

With tests = TRUE, a section under the plot, collapsed until the reader opens it, gives Egger's regression test (Egger et al. 1997), Begg's rank correlation test (Begg and Mazumdar 1994) and, with trim_fill = TRUE, the trim and fill estimate (Duval and Tweedie 2000), as computed by 'metafor' or 'meta', with a note on what each asks. Egger's test is the classical one, metafor::regtest(model = "lm"), which 'meta' also computes; metafor's own default, regtest() with model = "rma", gives a different p value. The two packages' trim and fill estimators can also impute different numbers of studies from the same data. The table is also returned as the field tests. These tests have little power with fewer than ten studies, and asymmetry can come from heterogeneity, chance or the quality of small studies as well as from publication bias.

With trim_fill = TRUE, the studies trim and fill imputes are drawn as hollow circles and the adjusted estimate as a dashed line, and the widget gets a switch that hides them.

Examples

if (requireNamespace("metafor", quietly = TRUE)) {
  dat <- metafor::escalc(measure = "RR", ai = tpos, bi = tneg,
                         ci = cpos, di = cneg, data = metadat::dat.bcg,
                         slab = paste(author, year))
  fit <- metafor::rma(yi, vi, data = dat)
  f <- ggfunnel(fit, hover = "alloc", trim_fill = TRUE)
  f
  f$tests
}
#>                      test
#> 1 Egger's regression test
#> 2 Begg's rank correlation
#> 3           Trim and fill
#>                                                    statistic         p
#> 1                                           t = -1.40, 11 df 0.1887070
#> 2                                         Kendall's τ = 0.03 0.9523619
#> 3 1 study imputed on the right; Risk ratio 0.52 (0.37, 0.74)        NA