Bias Correction Made Easy: A Web App for Meta-Analysis at EasyMeta.org
- Zuzana Irsova
- Sep 27
- 1 min read
Run MAIVE, PET-PEESE, and EK with one click — no coding, no installation.
We are happy to share a new tool now available at EasyMeta.org. The app makes the new MAIVE method (Meta-Analysis Instrumental Variable Estimator, published yesterday in Nature Communications) easy to apply with a few clicks. The site offers a demo dataset you can run instantly.
At the same time, the app allows for seamless use of methods well known in the MAER-Net community: PET-PEESE and the Endogenous Kink model. Until now, these approaches were available only in R or Stata. Now they can be run with a single click, no software installation needed.
The app supports options that applied researchers often need, including different types of clustering (classical, CR2, wild bootstrap), weighting schemes (including one that accounts for extreme heterogeneity), weak-instrument-robust confidence intervals, and fixed-intercept multilevel specifications to account for between-study differences in methods and quality.
We hope this tool will lower entry barriers to modern meta-analysis for students, collaborators, and applied researchers, and make it easier for MAER-Net members to demonstrate, compare, and teach bias-correction methods.
We’d be very grateful if you could use your favorite social network to share this tool with colleagues, students, or collaborators who might find it useful. And please let us know about any errors, missing features, or ideas for improvement — we’ll keep refining the app based on your feedback. Together we can make bias correction in meta-analysis more accessible.
Special thanks to Petr Čala for creating the app, and to Heiko Rachinger and Pedro Bom for developing the R package that powers it.
👉 Try a demo at EasyMeta.org/demo
Update (Nov 5, 2025): EasyMeta.org now also exports full R code used for the analysis, allows for study weights (giving each study, not each estimate, the same weight), and provides suggested interpretation.
We have incorporated most of the comments raised at the MAER-Net colloquium in Ottawa. The app now supports three model types:
1) WLS (standard PET-PEESE and Endogenous Kink)
2) MAIVE (the estimator from the Nature Communications paper, also available in inverse-variance weighted form)
3) WAIVE (a new, experimental extension of MAIVE that corrects more aggresively for p-hacking, https://meta-analysis.cz/waive_ottawa.pdf)
In addition, we offer the following options:
1) Anderson-Rubin confidence interval (recommended unless the first-stage F-stat is very large)
2) Standard error treatment (clustered CR1, CR2, wild bootstrap)
3) Even under MAIVE, you can choose different flavors (e.g. EK-MAIVE)
4) Different weights (equal, classical inverse variance, instrumented inverse variance)
5) Winsorization (0-5%)
6) Study intercepts/fixed effects (fixed-intercept multilevel), which is substantially more…