rvlenth
commented
1 year ago Thanks for reporting this. The error occurs because in your example, misc$tran is a list of length 7 when we are expecting a character value. The mystery to me is why this wasn't discovered long ago, since it will happen whenever the user uses make.tran() or make.link() and then does pairwise comparisons. I have corrected the code, and it will be in the next CRAN update.
Meanwhile, a workaround is to add ratios = FALSE) to the call:
> pairs(emms, ratios = FALSE)
contrast estimate SE df t.ratio p.value
A0 B0 - A1 B0 2.29 0.668 5 3.438 0.0644
A0 B0 - A0 B1 -0.53 0.668 5 -0.794 0.8547
A0 B0 - A1 B1 1.76 0.944 5 1.870 0.3451
A1 B0 - A0 B1 -2.83 0.944 5 -2.992 0.1024
A1 B0 - A1 B1 -0.53 0.668 5 -0.794 0.8547
A0 B1 - A1 B1 2.29 0.668 5 3.438 0.0644
Note: contrasts are still on the mu^0.1 scale
P value adjustment: tukey method for comparing a family of 4 estimates
Describe the bug
When calculating contrasts for an emmGrid object, I get an error:
"Error in ratios && true.con && misc$tran %in% c("log", "log2", "log10", : 'length = 7' in coercion to 'logical(1)'"
This only happens when I have specified a power transformation function to the emmGrid object.
To reproduce
I'm using R 4.3.0 and emmeans 1.8.7
require(pandas) require(emmeans) df = data.frame(resp=c(1,2,3,4,'0.12','0.17','0.23','0.3'),A=c(0,0,0,0,1,1,1,1),B=c(0,1,0,1,0,1,0,1)) fit = lm(resp~B+A,data=df) emms = emmeans(fit,data=df,~A+B,tran=make.tran(type=c("power"),0.1)) pairs(emms)
Expected behavior
I wanted to make pairwise comparisons for levels of my linear model. I added the transformation function so I could back-transform the estimates and confidence intervals. Seems that just the presence of power transform function causes the error, without even doing the back-transform.