You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Naive layperson food-for-thought question after skimming through some of the docs and codebase:
Could we make use of S3 (or S7 if we want to be forward-looking) classes/methods to improve ease of use (one of the roadmap items for mipdtrial), especially for cross-package functionality in our ecosystem?
As far as the API, maybe we could draw inspiration from tidymodels? (mainly thinking about the pipe-able, build a model in steps approach, which might work well for designing a trial too)
For context: I thought of this while reviewing PR #12, where we might want to have different approaches for different types of models. Using S3 classes might help us formalize how we handle situations like this, and make for clearer code and logic.
The text was updated successfully, but these errors were encountered:
Agree, making more robust and structured would be nice. Keep in mind though that the "model" concept does touch almost every other package we've created (PKPDsim, PKPDmap, irxanalytics, mipdtrial, PKPDposterior, ...), so there'd be a lot of code to update. If we do this, we should probably implement as optional structure first and keep supporting existing structure for the time being.
Naive layperson food-for-thought question after skimming through some of the docs and codebase:
Could we make use of S3 (or S7 if we want to be forward-looking) classes/methods to improve ease of use (one of the roadmap items for mipdtrial), especially for cross-package functionality in our ecosystem?
As far as the API, maybe we could draw inspiration from tidymodels? (mainly thinking about the pipe-able, build a model in steps approach, which might work well for designing a trial too)
For context: I thought of this while reviewing PR #12, where we might want to have different approaches for different types of models. Using S3 classes might help us formalize how we handle situations like this, and make for clearer code and logic.
The text was updated successfully, but these errors were encountered: