Skip to content

Change color in forest.rma (metafor)

11 messages · Bernd Weiss, Uwe Ligges, Paola Tellaroli +1 more

Paola Tellaroli
# 
My script is the following:

library(metafor)

yi<-c(-0.1, 0.2, 0.3, 0.4)
sei<-c(0.4, 0.2, 0.6, 0.1)
vi<-sei^2
studi<-c("A", "B", "C", "D")
eventi.c<-c(10, 5, 7, 6)
n.c<-c(11, 34, 25, 20)
eventi.a<-c(2, 7, 6, 5)
n.a<-c(11, 35, 25, 15)
dfs<-rma(yi, vi, method="DL")
dfs

windows(height=6, width=10, pointsize=10)
windowsFonts(B=windowsFont("Bookman Old Style"))

forest.rma(dfs, slab=studi, xlim=c(-15, 10), ilab=cbind(eventi.c, n.c,
eventi.a, n.a), ilab.xpos=c(-9.5, -8, -6, -4.5), cex=1.2, at=c(-2, -1, 0, 1,
2), family="B", xlab="Hazard Ratio (log scale)", mlab="Random Effects
Model", efac=5, col="red", border="red")
text(-10, -1.3, paste("Heterogeneity: I-squared=", paste(paste(round(dfs$I2,
2), "%", sep=""), paste("p", round(dfs$QEp, 4), sep="="), sep=", "),
sep=""), font=4, cex=1.2, family="B")

op<-par(cex=1.2, font=2, family="B", oma=c(0.5, 0.5, 0.5, 0.5), mar=c(0.5,
0.5, 0.5, 0.5))
text(x=c(-9.5, -8, -6, -4.5), 6, c("Events", "N", "Events", "N"), cex=1.2 )
text(c(-8.7, -5.5, 8), 6.5, c("S", "A", "Log"))
text(-15, 6, "Trials", pos=4)
text(10, 6, "Hazard Ratio [95% CI]", pos=2)
par(op)

Even if I have specified "col="red", border="red"", color of squares and
diamond rests black! Why?

Thanks,
Paola


--
View this message in context: http://r.789695.n4.nabble.com/Change-color-in-forest-rma-metafor-tp3765090p3765090.html
Sent from the R help mailing list archive at Nabble.com.
Bernd Weiss
# 
Am 24.08.2011 07:50, schrieb Paola Tellaroli:
As far as I know, "col" and "border" do only affect the fitted values
("diamonds"), i.e. the FEM/REM estimators (see ?forest.rma: "col:
character string specifying the name of a color to use for _the fitted_
values (?"darkgray"? by default).")

Furthermore, I had a quick look at the source code and it might be a
bug. If I replace in line 2770 the line

cex * efac), col = "black", ...)

with

cex * efac), col = col, ...)

you can at least specify your own colour. Changing the border color
seems a bit more tricky...

However, Wolfgang Viechbauer (the package author) is always a very
responsive and helpful person and I suggest you better wait for his answer.

Bernd
Viechtbauer Wolfgang (STAT)
# 
Thank you, Bernd, for looking into this. 

Yes, at the moment, the color of the summary estimate for models without moderators is hard-coded (as black). I didn't think people may want to change that. I guess I was wrong =)

A dirty solution for the moment is to add:

addpoly(dfs, efac=6, row=-1, col="red", border="red", annotate=F, mlab="")

after the call to forest(). You will get a warning message (since the border argument gets passed to the text() function inside addpoly() and that's not a par for text), but you can just ignore that.

Best,
Viechtbauer Wolfgang (STAT)
# 
The color of the squares is also currently hard coded. 

The thing is, there are so many different elements to a forest plot (squares, lines, polygons, text, axes, axis labels, etc.), if I would add arguments to set the color of each element, things would really get out of hand (as far as I am concerned, there are already too many arguments to begin with). I can think of one possibility: I could allow the col argument to accept a vector of colors and then apply the different elements of that vector to the various elements in the plot. Of course, there is also a limit to how far that can be taken. For example, what if somebody wants to have a different color for *one* of the squares and a different color for the other squares?

Another possibility is to do some post-processing with other software. One can create the forest plot in R, save it for example as a postscript file, and the edit the plot in other software. Yes, I prefer it if I can create the plot in R and have it exactly the way I want it (without having to do any post-processing), but sometimes that may not be possible.

Note that you can always add whatever you want to a plot created by the forest() function after it has been drawn. You can add text, lines, squares, polygons, whatever in any color you desire (e.g., with the text(), segments(), points(), polygon() functions). So, you could also just plot over the squares with:

points(yi, 4:1, pch=15, col="red")

To get rid of the black squares that are drawn by the forest function, add psize=0 as an argument in forest() (this will make the size of squares equal to 0, so essentially, they are invisible).

If you want to make the size of the points inversely proportional to some function of the precision of the estimates, use points() together with the cex argument. For example:

wi <- 1/sqrt(vi)
psize <- wi/sum(wi)
psize <- (psize - min(psize)) / (max(psize) - min(psize))
psize <- (psize * 1.0) + 0.5
points(yi, 4:1, pch=15, col="red", cex=psize)

Best,

Wolfgang
2 days later
Uwe Ligges
#
On 26.08.2011 15:50, Paola Tellaroli wrote:
You can, see ?par and its "xpd" argument.

Uwe Ligges
Viechtbauer Wolfgang (STAT)
# 
See ?par and its "mar" argument.
Uwe Ligges
#
On 29.08.2011 13:11, Paola Tellaroli wrote:
Please read the whole help page for ?par

You will find a way how to increase the size of the margins (using the 
argument "mar") and many other useful things.

Uwe Ligges