Function to examine publication bias. For both fixed- and random-effects models,
estimates from no-pooling effects model are used as study-specific estimates. For
random-effects models, the corresponding fixed-effects models are implemented at
background to obtain pooled estimate. For example, if users call bmeta to
run random-effects meta-analysis with normal prior, fixed-effects meta-analysis
with normal prior are implemented at background to obtain pooled estimate for
graphing. In the absence of publication and heterogeneity, the scatter resembles
a symmetrical funnel and the triangle area formed by connecting the centred summary
estimate with its 2.5% and 97.5% quantiles on either side includes about 95% of
the studies if the fixed-effects model assumption holds (i.e. all the studies
estimate the same effect).
title of x-axis. If unspecified, the function sets an appropriate lable by default.
ylab
title of x-axis. If unspecified, the function sets an appropriate lable by default.
title
title of the plot if specified
xlim
horozontal limits of the plot region. If unspecified, the function sets the horizontal
plot limits to (-6,6).
Author(s)
Tao Ding
Gianluca Baio
Examples
### Read and format the data (binary)
data = read.csv(url("http://www.statistica.it/gianluca/bmeta/Data-bin.csv"))
### List data for binary outcome
data.list <- list(y0=data$y0,y1=data$y1,n0=data$n0,n1=data$n1)
### Select random-effects meta-analysis with t-distribution prior for binary
### data
x <- bmeta(data.list, outcome="bin", model="std.dt", type="ran")
### using output from bmeta to produce funnel plot
funnel.plot(x)
### using output from bmeta and specify title of the plot
funnel.plot(x,title="funnel plot")
### using output from bmeta and specify the limit of x-axis and title
funnel.plot(x,title="funnel plot",xlim=c(-2,1))