Open vincentarelbundock opened 2 months ago
Hi!
I am a fan of the work you've done with marginaleffects
! I remember the old marginaleffects::deltamethod()
, which has been replaced by hypotheses()
. The main difference is that tidydelta()
can take a tibble or data frame, while hypotheses()
is focused on model objects. In comparison to car::deltaMethod()
, it provides more flexibility in providing inputs.
As for avoiding duplication of efforts, it is a coding and package-making exercise. Since I've already put in the work, I thought of sharing it in case it could be helpful for anyone looking for a Delta Method with these features.
I remember the old
marginaleffects::deltamethod()
, which has been replaced byhypotheses()
.
Cool cool.
As for avoiding duplication of efforts, it is a coding and package-making exercise. Since I've already put in the work, I thought of sharing it in case it could be helpful for anyone looking for a Delta Method with these features.
Love this attitude!
The main difference is that
tidydelta()
can take a tibble or data frame, whilehypotheses()
is focused on model objects.
FYI, marginaleffects::hypotheses()
also takes a tibble or data frame. See example here: https://marginaleffects.com/vignettes/hypothesis.html#arbitrary-quantities
FYI, marginaleffects::hypotheses() also takes a tibble or data frame. See example here: https://marginaleffects.com/vignettes/hypothesis.html#arbitrary-quantities
Thanks for bringing this up. I wasn't aware of this approach to using marginaleffects::hypotheses()
! 🙌🏽
Hi,
I just noticed.the CRAN release. Congratulations!
In case this can prevent duplication of efforts, I thought I'd point out that the
marginaleffects
package has a simpler (but much more flexible) implementation of the delta method.You might be in interested in the
hypotheses()
function described in this vignette: https://marginaleffects.com/vignettes/hypothesis.htmlAlso note that the outputs of that package are also perfectly "tidy": they are simple data frames which follow the
broom
column naming conventions.