metagroup
provides a suite of tools to uncover hidden
structures in meta-analytic data. It uses a two-step process to perform
meaningful subgroup analysis:
mgbin()
, mgcont()
) to partition studies into
statistically homogeneous clusters based on their effect size data.meaning()
function
and its associated plot()
method to analyze these new
subgroups and understand their composition based on study-level
characteristics (e.g., country, setting).This approach helps to provide a deeper, more data-driven interpretation of heterogeneity in a meta-analysis.
You can install the development version of metagroup
from GitHub with:
# install.packages("remotes")
::install_github("asmpro7/metagroup") remotes
Here is a basic example of the core workflow: first grouping the studies, then finding the meaning behind the groups.
# 1. Load the package
library(metagroup)
# 2. Step 1: Group the studies by homogeneity
# The result contains the original data with a new 'subgroup' column
<- mgbin(
grouped_results data = study_data,
event.e = event.e,
n.e = n.e,
event.c = event.c,
n.c = n.c,
studlab = studlab,
sm = "OR"
)
# 3. Step 2: Analyze the composition of the new subgroups
<- meaning(
meaning_results data = grouped_results,
variables = c("country", "setting")
)
# Print the summary table to see the dominant characteristics of each group
print(meaning_results)
# Plot the results to visualize the composition of all groups
plot(meaning_results)
This package is licensed under the GPL-3 License.