Closed bbolker closed 7 years ago
I'm amenable to that. This is code that was refactored out of margins for maintainability. My impression is broom doesn't implement a predicted/fitted values for particularly large range of model classes does it? (I haven't used it much, so I don't really know.)
(FWIW I haven't even asked the broom maintainer about this yet.) This might nicely augment the current 'verbs' in broom, which are tidy
(coefficient table), augment
(original data + residuals, predictions, etc. -> update of old ggplot2::fortify
) and glance
(model summaries such as AIC, R2, etc.). augment
is really designed around the original data (although e.g. augment.lm
does take a newdata
argument ...)
Ah, okay. For this, being able to pass new data is kind of essential. I'll take a more thorough look at broom at some point and see how this might all fit together.
Radical thought, but would you consider/what would you think about merging this functionality with the
broom
package? It feels like it would be easier to have a smaller number of packages for dealing with tidy model outputs ...