Frequentist and Bayesian Meta-Analysis of a Sarcoma Prognostic Model

This repository contains the code and data used for the meta-analysis conducted as part of my Master's thesis in the MSc Applied Statistics programme. The objective was to evaluate the performance of a prognostic model for sarcomas using both frequentist and Bayesian methods. The code in this respository can be used to run the frequentist and Bayesian meta-analysis, construct forest plots, conduct a sensitivity analysis by excluding a study with high risk of bias and plot a bar chart and traffic light chart of the risk of bias.

Repository structure

.
├── data/
│   ├── dat_ma.csv   # Data for the meta-analysis
│   └── dat_rob.csv  # Data for the risk of bias plots
├── meta_analysis.R      # Runs frequentist and Bayesian meta-analyses
├── forest_plot.R        # Forest plot of the full analysis
├── forest_plot_sens.R   # Forest plot excluding the study with high bias
└── risk_of_bias.R       # Bar chart and traffic light plot of risk of bias

Usage

1. Run the meta-analysis Execute meta_analysis.R first. This script loads dat_ma.csv and performs both frequentist and Bayesian random-effects meta-analyses using the metamisc package.

2. Display forest plots

   After running the analysis, use forest_plot.R to visualise the full results and forest_plot_sens.R to view the sensitivity analysis that excludes the study with high risk of bias.

3. Assess risk of bias

   Run risk_of_bias.R to create a bar chart and traffic light plot summarising the risk-of-bias assessments stored in dat_rob.csv.

The scripts rely on the packages metamisc, coda, ggplot2, and dplyr. Install them in your R environment if they are not already available. In order to perform the Bayesian meta-analysis, the metamisc package depends on the packages runjags and rjags. JAGS must be installed on the local machine to run the rjags package.