Meta-Analysis Resources

Tools for Those Who Summarize the Evidence Base

Resources and networking for those who conduct or interpret meta-analyses related to any phenomenon that is gauged in multiple studies.

Graphical displays and point estimates with meta-regression

I'm curious: Have any members of this site seen any good treatises on the subject of graphing meta-regression results or determining what the confidence interval is at a certain point along a regression line? Or, in terms of plotting meta-regression lines with confidence bands?

My impression is that the classic, most-used, and even new books on meta-analysis always describe how to get a mean and a confidence interval but never how to see what it looks like across a range of an independent variable, or at a predicted point along the independent variable. Can you think of any exceptions?

(As side note, this site has a group related to such issues, the one focused on analyses). (Also, students in my most recent meta-analysis seminar will realize that Tania Huedo-Medina and I have been working on a paper on this subject. I just want to see what the collective wisdom is.)

THANKS. -Blair

Views: 163

Replies are closed for this discussion.

Replies to This Discussion

Most books on meta-analysis skim around this issue, since one essentially has to use matrix algebra notation to write down the required equations. However, if I remember correctly, the chapter by Raudenbush in the Handbook should have them.

A plot of what you have in mind can be found in my paper on the metafor package (http://www.jstatsoft.org/v36/i03/). In particular, see Figure 2. It shows the estimated/predicted relative risk as a function of absolute latitude for the BCG vaccine data. The corresponding 95% confidence interval bounds are also added to the plot.

I think figures like this are quite underused in meta-analyses. Many people would end up dichotomizing the moderator (i.e., close to equator, far from equator) and fit separate random-effects models within each level.
Thanks Wolfgang!
All, we've prepared a manuscript detailing the strategy and providing examples and code in two major platforms (R and Stata). If you are interested to see it, please send me an email.