Skip to main content

Doing Meta-Analysis in R: A Hands-on Guide

E Corrections & Remarks

Errata and remarks concerning the first edition print version of the book are displayed here.

Last updated: 09 August, 2024.

Chapter 3.3.3

In the imaginary experiment in Figure 3.4, four participants experience the event during the 10-year observation period. In the calculation example, however, we use \(E=3\) to calculate the incidence rate. In the online version, we have now corrected this so that \(E=4\), like in the experiment.

Chapter 4.2

A new version of {meta} (version 5.0-0) has recently been released. We adapted the code in this chapter accordingly to avoid deprecation messages:

  • The comb.fixed and comb.random arguments are now called fixed and random, respectively.

  • To print all studies, one now has to use summary method for {meta} meta-analysis objects.

In the latest version of {meta} (version 7.0-0), calls to update.meta now need to be replaced with update. If meta models have been fitted using the deprecated hakn argument, update will now produce an error message. We have therefore changed this to the new method.random.ci argument throughout the guide. This applies to all meta-analysis functions, e.g., metagen, metabin, metacor, etc.

Chapter 7.3

A new version of {meta} (version 5.0-0) has recently been released. We adapted the code in this chapter accordingly to avoid deprecation messages:

  • The byvar argument is now called subgroup.

  • To print all studies, one now has to use summary method for {meta} meta-analysis objects.

Chapter 12.2.1

The print version contains a factual error concerning the definition of full rank in non-square (rectangular) matrices. It is stated that a “matrix is not of full rank when its rows are not all independent”. This, however, only applies to square matrices and non-square matrices with less rows than columns (\(m < n\)). In our example, there are more rows than columns; this means that \(\boldsymbol{X}\) is not full rank because its columns are not all independent (in \(m > n\) matrices, rows are always linearly dependent). This erratum has been corrected in the online version.

Chapter 12.2.2

A new version of {netmeta} (version 2.0-0) has recently been released. We adapted the code in this chapter accordingly to avoid error messages:

  • The latest version of {netmeta} resulted in non-convergence of the Fisher scoring algorithm implemented in rma.mv. This problem pertains to all versions of {dmetar} installed before 24-Oct-2021. To avoid the issue, simply reinstall the lastest version of {dmetar}.