ma-cc-rare

Code for The Impact of Corrections Methods on Rare-Event Meta-Analysis by Brinley N. Zabriskie, Nolan Cole, Jacob Baldauf, and Craig Decker.

This folder contains code to reproduce the simulation study (Section 4 in the paper) and the case study example (Section 5 in the paper).

For the simulation study:

Create two directories: RDSDataSets and Results. Make sure the following packages are installed: meta, metafor, BiasedUrn, future.apply, and nleqslv.

0 Miscellaneous scripts needed for subsequent scripts (do not run):

• 0_DGMs.R
  • Contains binary event meta-analysis data generating functions
• 0_CCCode.R
  • Contains code to implement various continuity corrections to individual MA data sets
  • Calls 0_MAResults.R
• 0_ApplyAllCCs.R
  • Applies various continuity corrections to individual MA data sets
  • Calls 0_CCCode.R
• 0_MAResults.R
  • Runs an individual MA data set through metabin/metafor, separated by heterogeneity variance estimator
  • Calls 0_IPM.R
• 0_IPM.R
  • Contains code to implement the improved Paule-Mandel (IPM) heterogeneity variance estimator

1 Scripts to generate meta-analysis data:

• 1_GenerateMAData.R
  • Generates binary event meta-analysis data
  • Data generated is stored in the folder RDSDataSets
  • Need to set the appropriate amount of workers
  • Calls 0_DGMs.R
• 1_GetpCFixedResults.R
  • Applies various continuity corrections, heterogeneity estimators, and pooling methods to previously generated data sets
  • This code is for pCFixed – do the same things for pRandomB
  • Need to set the appropriate amount of workers
  • Calls 0_ApplyAllCCs.R and 0_MAResults.R
• 1_GetpCFixedGLMMResults.R
  • Applies the GLMM to previously generated data sets
  • This code is for pCFixed – do the same things for pRandomB
  • Need to set the appropriate amount of workers Results can then be summarized in terms of Type I error rate, relative power, confidence interval coverage, median confidence interval width, mean squared error, and median bias, as outlined in the paper.

For the case study example:

Make sure the following packages are installed: tidyverse, readxl, and meta.

1. Anti-TNF.xlsx
   • Anti-TNF data set
2. Case_Study_Anti-TNF.R
   • Applies various continuity corrections, heterogeneity estimators, and pooling methods to the Anti-TNF data set
   • Reads in Anti-TNF.xlsx
   • Calls 0_ApplyAllCCs.R and 0_MAResults.R (defined above)