# Meta analysis: transform forest graph output to percentage effect in R {metafor}

I need the forest graph expressed as % effect. I have run the model with the log ratio, and I know I could back-transform using transf=exp, which would plot the forest graphs expressed as the ratio (r). What I want though is the percentage (r-1)*100.

   > effectSite
SITE SPECIES  CO2  N       NPP        sd         se        ci
1         DUKE    PITA  amb 24  982.3750 173.75090  35.466754  73.36857
2         DUKE    PITA elev 24 1285.5000 224.99256  45.926414  95.00603
3         ORNL      SG  amb 33  788.5616 164.14384  28.573776  58.20288
4         ORNL      SG elev 22  909.1108 207.28113  44.192485  91.90330
5      POPFACE    alba  amb  6 1312.5000 207.62635  84.763101 217.89049
6      POPFACE    alba elev  6 1685.3333 484.02300 197.601563 507.95099
7      POPFACE     eur  amb  6 1331.1667 236.40678  96.512665 248.09370
8      POPFACE     eur elev  6 1691.1667 232.79383  95.037683 244.30214
9      POPFACE   nigra  amb  6 1710.3333 301.22860 122.976059 316.12002
10     POPFACE   nigra elev  6 1987.5000 480.73392 196.258800 504.49931
11 Rhinelander      AA  amb  3  424.0000  13.00000   7.505553  32.29379
12 Rhinelander      AA elev  3  541.3333 106.02515  61.213651 263.38108
13 Rhinelander      AB  amb  9  283.1111  45.27815  15.092718  34.80387
14 Rhinelander      AB elev  9  486.3333  93.55881  31.186268  71.91566

effectSite$CO2 <- factor(effectSite$CO2,levels=c("elev","amb"))
require(metafor)
dat <- escalc(measure="ROM",data=effectSite,
formula=(NPP/sd) ~ CO2 | SPECIES,weights=N) # ROM= log ratio of means
FEmodel <- rma(yi,vi,data=dat,method="FE")
forest(FEmodel)


Just write your own transformation function and use that. If I understand you correctly, you want:

mytransf <- function(x)
(exp(x) - 1) * 100


Then you can use:

predict(FEmodel, transf=mytransf)


And:

forest(FEmodel, atransf=mytransf)


Or:

forest(FEmodel, transf=mytransf)


The former only transforms the axis/annotations, while the latter transforms the values directly. See help(forest.rma) for the difference between the transf and atransf arguments in the context of drawing forest plots.