Looking over #52 got me revisiting package and repo scope.
It is hard to disentangle evaluation and fitting, but a "package for evaluable lineage modeling" is possible, which would have standards for the data you can put in but wouldn't be able to create it. That is, it would require "clean data" from the flow chart and remove linmod.data.
On the other hand, a package that is only the modeling and evaluation (and perhaps plotting thereof) isn't directly usable. Having NextStrain interfacing baked in has been a big boon for making progress in development.
As it is, we already have pipeline folders and the linmod package/folder in this repo, so rather than carving out linmod into an entirely separate repo, or removing the non-linmod components, we could data-wrangling to its own one-script "package" (data or nextstrain-data).
Looking over #52 got me revisiting package and repo scope.
It is hard to disentangle evaluation and fitting, but a "package for evaluable lineage modeling" is possible, which would have standards for the data you can put in but wouldn't be able to create it. That is, it would require "clean data" from the flow chart and remove
linmod.data.On the other hand, a package that is only the modeling and evaluation (and perhaps plotting thereof) isn't directly usable. Having NextStrain interfacing baked in has been a big boon for making progress in development.
As it is, we already have pipeline folders and the
linmodpackage/folder in this repo, so rather than carving outlinmodinto an entirely separate repo, or removing the non-linmodcomponents, we could data-wrangling to its own one-script "package" (dataornextstrain-data).