13: Forest PlotsMaking them look the way you want

Meta-analysis in R with Metafor

Forest Plots 2

The forest plot may not look the way you want by default.

Good News: Using R, you can make it look most any way you want; it’s very flexible.

Bad News: Flexibility and ease of use are inversely related.

I know little graphics in R, so I will only show some basics that you can use in preparing your forest plot for publication.

There are tutorials on graphics in R, for example:http://ww2.coastal.edu/kingw/statistics/R-tutorials/graphs.html

You can output R graphics and edit them with other programs (e.g., PowerPoint, Inkscape).

R CodeI am working with an epidemiologist on a meta-analysis

The effect sizes are odds ratios; analysis was done in log(odds)

We want to portray the distribution of ES and show a moderator (kinds of sleep disorder)

Moderator test, sorted by moderator and ES (k=48)

Split into 2 smaller samples (select last 2 from full sample, k=23)

Add study labels (slab=…)

Convert to odds from log(odds) (atransf=…)

Change ES label from observed to odds (xlab =…)

Make scales the same for both graphs (at=…)

Add labels to top of figure (text…)

Add labels to groups (text…)