TimelyScholarFree Tools › Meta-Regression Data Builder
Lay out study-level moderators alongside effect sizes and standard errors, then download a clean CSV and ready-to-run code for R (metafor) and Stata.
FreeNo account neededYour data stays in your browser
| Study | Effect (yi) | SE | latitude | allocation | |
|---|---|---|---|---|---|
6 studies entered. The Cochrane Handbook advises that meta-regression should generally not be considered with fewer than ten studies.
library(metafor)
dat <- read.csv("meta-regression-data.csv")
# random-effects meta-regression (REML)
res <- rma(yi = yi, sei = sei, mods = ~ latitude + factor(allocation), data = dat, method = "REML")
summary(res)
# bubble plot for the first moderator
regplot(res, mod = "latitude")import delimited "meta-regression-data.csv", clear meta set yi sei, studylabel(study) meta regress latitude i.allocation
For Comprehensive Meta-Analysis (CMA), import the CSV and set the effect and standard error columns, then add the moderators as study-level covariates. Effects should be on the analysis scale (for example log odds ratios).
Mark each one as numeric or categorical.
Effect size on the analysis scale, its standard error and the moderator values.
Get the CSV plus R and Stata code for a random-effects meta-regression.
Meta-regression examines whether study characteristics (moderators) such as dose, latitude or risk of bias explain differences between study results. Each study contributes one effect size, its standard error and its moderator values.
The Cochrane Handbook advises that meta-regression should generally not be considered when there are fewer than ten studies, and that moderators should be specified in advance to limit false-positive findings.
The analysis scale: log odds ratios or log risk ratios for ratio measures, and raw values for mean differences or standardised mean differences.
Few. With a small number of studies, each extra moderator reduces power and increases the chance of spurious findings.