Feature Request: Regression examples for ard_regression

[daniel-woodie]

What's the feature?

The example in the documentation is really just scratching the surface of what's possible. I'd like to use these capabilities but I'm not sure where to start with, say, how to run an mmrm in cardx. Happy to work on a pull request if that's easier here.

[ddsjoberg]

Thanks for the post @daniel-woodie !

I just took a look at the help file, and it is indeed pretty sparse 😬

Perhaps we could add a section where we list the supported models broom.helpers::supported_models$model

betareg::betareg() biglm::bigglm() brms::brm() cmprsk::crr() fixest::feglm() fixest::femlm() fixest::feNmlm() fixest::feols() gam::gam() geepack::geeglm() glmmTMB::glmmTMB() glmtoolbox::glmgee() lavaan::lavaan() lfe::felm() lme4::glmer.nb() lme4::glmer() lme4::lmer() logitr::logitr() MASS::glm.nb() MASS::polr() mgcv::gam() mice::mira mmrm::mmrm() multgee::nomLORgee() multgee::ordLORgee() nnet::multinom() ordinal::clm() ordinal::clmm() parsnip::model_fit plm::plm() pscl::hurdle() pscl::zeroinfl() rstanarm::stan_glm() stats::aov() stats::glm() stats::lm() stats::nls() survey::svycoxph() survey::svyglm() survey::svyolr() survival::cch() survival::clogit() survival::coxph() survival::survreg() tidycmprsk::crr() VGAM::vgam() VGAM::vglm()

And another section highlighting/describing the goodies we get from the package.

I am hesitant to add an mmrm example, because that will add an mmrm dependency to the package and that just complicates maintenance. Is there any special results from mmrm that can't be illustrated with glm()?

[daniel-woodie]

Thanks for the reply! Okay. Interesting. So these models are supported but not included in the package? Makes sense but also not obvious. Are they supported because they provide a list response that can be ingested by broom? Or is there more to it than that?

I think including that list would be helpful. It makes sense not to bloat an R package with these methods... it also makes sense haha. Ultimately, your call here and definitely understand why you wouldn't want to. One pro that comes to mind is it eases the eventual 'validation' -- well, it at least centralizes what needs to be validated (e.g. if {cards} and {cardx} are considered validated by your org then you're good to go). ... this argument works both ways, though, and makes the validation burden higher.

Anyway, yes, including that list of functions would be helpful. Additionally, if you don't want to include the example on that page then it might be good to have a reference article going over how to leverage one of these regression models. For example, I didn't realize mmrm was supported so I built my own version and just fed it in as a custom function into ard_complex() using the statistic parameter. I haven't tested out how to do this through ard_regression() but that's what I'm wondering about now. Additionally, if I wanted to limit the parameters which are included in the ARD, how would I do that?

[daniel-woodie]

Side note. I attended the workshop at phuse and you mentioned some companies had formally partnered with you on {cards}/{cardx} (correct me if I heard that incorrectly). I've been testing the waters through GitHub issues but happy to talk more about formal involvement/partnerships from that entry point.

[ddsjoberg]

Side note. I attended the workshop at phuse and you mentioned some companies had formally partnered with you on {cards}/{cardx} (correct me if I heard that incorrectly). I've been testing the waters through GitHub issues but happy to talk more about formal involvement/partnerships from that entry point.

YESSS! That would be awesome! I'll send you an message to connect.

[ddsjoberg]

Regarding your other message. The mmrm package is a dependency on broom.helpers and has unit tests there to ensure continued compatibility as that package continues to develop FYI

Would you like to submit a PR updating the help file for ard_regression() with more details? After we have that in place, we can revisit the convo about adding mmrm examples.