Log transform is not detected when applying emmeans to a list of models

[qdread]

formula_string <- 'log(mpg) ~ factor(cyl)'
model <- lm(as.formula(formula_string), data = mtcars)
emmeans(model, specs = ~ cyl, type = 'response')

formula_string_list <- list('log(mpg) ~ factor(cyl)', 'log(disp) ~ factor(cyl)')
models <- lapply(formula_string_list, function(form) lm(as.formula(form), data = mtcars))
lapply(models, function(mod) emmeans(mod, specs = ~ cyl, type = 'response'))

In the above, I get the expected output with back-transformed emmeans from the first call to emmeans(), where the formula is coerced from a string. However in the second case, where the lm() function is applied to a list of strings coerced to formulas, the log transformation is not detected and the means are not back-transformed. I tried to work around this using regrid() and summary(type = 'r') but could not find a workable solution. Thanks in advance for your help!

[rvlenth]

Tracing this for models, I get the error message:

Error in eval(mf, parent.frame()) : object 'form' not found

That is, we have lost the environment of your function call. I modified the code so the evaluation is done in environment(trms) where trms is based on the terms() component, and it seems to work now:

> ref_grid(model)
'emmGrid' object with variables:
    cyl = 4, 6, 8
Transformation: “log” 

> ref_grid(models[[1]])
'emmGrid' object with variables:
    cyl = 4, 6, 8
Transformation: “log” 

So it is fixed in the next push to GitHub, as long as this fix doesn't break something else...

[rvlenth]

The workable solution by the way is to explicitly specify the transformation:

lapply(models, function(mod) emmeans(mod, specs = ~ cyl, type = 'response', tran = 'log'))

[qdread]

Thank you for tracking down the source of the problem, fixing it, and providing a more robust solution! What more could one ask for

[qdread]

Oops I did not mean to close the issue

[rvlenth]

OK. But if you're not closing, does that mean that there's more to do to address the issue? If so, I'm not seeing it.

[qdread]

No it is all good! Thanks for addressing. Closed again ;-)