fix issue 658 by reproducing formula function

[iago-pssjd]

Close issue #658. The idea, as told there, is to do fm <- formula(x$terms) instead of fm <- eval(x$call$fixed) , but this would need to include importFrom("stats", formula) in the NAMESPACE, and since there is no importFrom there, I assumed you prefer not use them. Therefore, instead of using formula(....), I wrote the code of stats:::formula.terms there.

[vincentarelbundock]

Thanks for opening a pull request!

You don't need to use importFrom or to re-write the code yourself. Instead, you can simply call the function from its namespace:

fm <- stats::formula(x$terms)

Are we sure that the two versions will always give the same results? What if there are variables in both the random and fixed components/formulas? Do both codes give the same results in a model estimated outside a lapply?

I'm not sure what the original intent of the code was in that respect, so it would be important to make sure nothing changes in different models.

[iago-pssjd]

@vincentarelbundock thanks for the point of using stats::formula. I believe this is more a stylistic which you, the package developers, know better than me your preferences.

Are we sure that the two versions will always give the same results? What if there are variables in both the random and fixed components/formulas? Do both codes give the same results in a model estimated outside a lapply?

I'm not sure what the original intent of the code was in that respect, so it would be important to make sure nothing changes in different models.

It seems so, at least with the original formulas. I just have built, installed and checked --as-cran the package without any complaints. It lacks a revdep check, so it does not changes other packages behaviours, but I am not sure how and do not have enough time to do it.

Update: I forgot to say that I had looked for the details of the terms element but it seems to be undocumented (?lmeObject)

[iago-pssjd]

You don't need to use importFrom or to re-write the code yourself. Instead, you can simply call the function from its namespace

I updated the code using now stats::formula.

[strengejacke]

Do we have some example code where the former code failed, and your PR works? We could then also use this demonstration to add some tests.

[iago-pssjd]

Yes! Issue #658:

library(insight)
library(nlme)
models <-lapply(c("", " + Sex"), \(.x) lme(as.formula(paste0("distance  ~ age", .x)), random = ~ 1, data = Orthodont))
lapply(models, find_formula)

Or just find_formula(models[[1]])

[vincentarelbundock]

Thanks a lot for the PR!

I merged it and added a test here: #660