I have created a simple linear regression model in R where I have log transformed the response. The independent variable is categorical with 80 levels. Then I use emmeans for pairwise tests. How can I backtransform the emmeans to have the same scale as the response (not the log scale)? I use the emmeans package in R.
I think this is just
summary(emmeans(...), type = "response")
? There is a vignette on transformations and link functions
e.g.
library(emmeans)
m1 <- lm(log(mpg) ~ factor(cyl), mtcars)
summary(emmeans(m1, ~cyl), type = "response")
cyl response SE df lower.CL upper.CL
4 26.3 1.268 29 23.9 29.0
6 19.7 1.190 29 17.4 22.3
8 14.9 0.636 29 13.6 16.2
Confidence level used: 0.95
Intervals are back-transformed from the log scale
Note that if you use pairs(emmeans(m1, ~cyl), type = "response") it will give you the pairwise ratios, but notes that the tests are performed on the log scale (as they should be)