Vague: opinionated modelling? - Githubissues

alexpghayes / modelling-in-r

an initial attempt to describe a grammar of modelling for r

https://alexpghayes.github.io/modelling-in-r/

3 stars 0 forks source link

Vague: opinionated modelling? #17

Closed alexpghayes closed 6 years ago

alexpghayes commented 6 years ago

Hadley attributes a bunch of the success of tidyverse to opinionated interfaces that prevented users from doing silly things and made it easy for them to do smart things.

I think this interface will make it easier to do smart things. Are there places were we can make it harder to do silly things?

What are "silly things" in terms of modelling? Need someone with a bunch more background here, but the first things that spring to mind are:

data leakage problems / poor hyperparameter validation / never assessing out of sample error
hitting problems blindly with lots of models and getting stuck after that

Can data leakage be at least partially solved by smart fit defaults?

alexpghayes commented 6 years ago

Closing to clean up issues and because I think this will be resolved by:

sane default interface to fit that specifies principled resampling
a standardized set of methods for probing models after they've been fit