Detecting Publication Bias in Meta-Analysis

This project investigates statistical methods for detecting and addressing publication bias in meta-analysis.
It illustrates applied statistical computing in R, combining methodological rigor with reproducible workflows.

---

Objectives

• Explore methods to identify bias in published research
• Apply regression-based tests (e.g., Egger’s regression)
• Model selection mechanisms that drive bias
• Demonstrate corrections using state-of-the-art R packages
• Communicate findings with reproducible reports

---

Methods and Tools

• R Programming for statistical analysis and reproducibility
• RMarkdown/Quarto for integrated code and narrative reporting
• Core Packages:
  • metafor: effect size computation, random- and mixed-effects models
  • clubSandwich: robust variance estimation
  • weightr / selection.model: publication bias modeling
  • ggplot2: funnel plots, forest plots, and diagnostics

---

Example Analyses

• Funnel plots to visualize asymmetry
• Egger’s regression test for small-study effects
• Selection models for adjusting bias
• Robust variance estimation for sensitivity analysis

---

Reproducibility

The analysis is contained in RMarkdown, ensuring that results are transparent and reproducible.
To reproduce:

# Install required packages
install.packages(c("metafor", "clubSandwich", "ggplot2"))
# For bias modeling
install.packages("weightr")

# Render the report
rmarkdown::render("Detecting-PB.Rmd")