Hello!
I want to create a forest plot with subgroups. The goal is to combine various effect sizes from different factors with subgroups into one effect size, similar to the attached figure. For example, I would like to combine all the effect sizes corresponding to Canada into one size effect rather than having all the observations for Canada plotted individually, as shown in the attached figure. The same is for Germany, the USA, Sweden, etc. Once the individual effect sizes for these countries are grouped, they should be placed under the label Country.
The code that I typed into R to create the plot that I am presenting is as follows:
dat
This is a lot easier to do using meta instead of metafor. Meta is essentially a wrapper for metafor anyway, and you can obtain identical results by feeding your effect sizes from escalc to the generic metagen() function and specifying the maximum likelihood estimator for variance. You can use the byvar argument in metagen to add subgroups, and then all of the options you are looking for are available in the meta::forest function.
If you want a plot that looks identical to the one you have shown, you can use the forester package (https://github.com/rdboyes/forester).
If you absolutely have to use metafor:
You can use the cex argument to increase the scaling factor of the font, which will also apply to the point sizes unless they are fixed with the psize argument.
The easiest way to plot only the subgroup results as individual results it to export them to a vector and plot the vector using forest.default(). To do this, you will have to first create rma objects for each group i.e.
rma.canada
Jack Henry, I hope all is well. I have several group effects: country, breed, method, season, handling, floor, ventilation, bedding, removal, feed.system, and feed.frequency. To ensure I get the code right, I started only with the effect of country, so I typed the following code:
dat
Apologies - try
forest.default(x = forest$TE, sei = forest$sei, slab = forest$slab)
Hi Jack Henry ,
I hope all is well! The code you kindly shared with me to construct the forest plot worked wonders; however, I must improve the image by including sub-groups, the overall effect size, labels, and heterogeneity. I have been working on these improvements but have not been successful. Could you help me with these changes, please?
So far, the code I have to build the plot is as follows:
dat
Alejandro Castaneda Without knowing what subgroups you are trying to create, etc., I can't be very specific. I would again recommend using meta rather than metafor to create this plot. If you do use metafor, here is an example of how to create such a plot: https://www.metafor-project.org/doku.php/plots:forest_plot_with_subgroups.
The general process is:
##add grouping variable to data frame with estimates
library(metafor)
library(tidyverse)
forest$group
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Meta-Analysis