Section IV

[dill]

Okay, I've just gone through my edits and comments on section IV (see 7f01d763877ed53a1d371f254745480977bcf3b2) and I'm wondering about what to do with this section in general.

I think the paper is rather long, and this section has some very useful stuff in it (but that stuff tends to be a bit technical), but I wonder if there are places we can chop bits out and "appendicize" a fair proportion of it, so the casual read who wants to know how to get going with factor-smooth interactions can get right from the model setups to the examples?

So here is a list, subsection-by-subsection with my thoughts...

• "Bias-variance tradeoffs": I think the first two paragraphs can be dropped/integrated into the following text. Probably this issue can be alluded to in the text then appendicized.
• "Complexity--computation trade-offs": again, this has a lot of useful info in it, but I'm not sure this should be in the main body of the paper in its current form. Again, I think these issues can be alluded to briefly in the main text, then ellaborated in an appendix.
• "Alternative formulations: bam, gamm, and gamm4": I think this would make a great appendix on more complicated computational issues
• "Estimation issues when fitting both global and groupwise smooths": seems like this again would be a useful issue to touch on more briefly then flesh out in an appendix.
• "A brief foray into the land of Bayes": I cut this fairly brutally (sorry; I'm pretty much fully on the Bayes boat now and I think we've been Bayesians all along, see edits), I'd rather see the generic theoretical content integrated into the text from the start if possible and then have the brms example appendicized (I've commented out this last bit for now).
• "Beyond HGAMs: functional regression": I think this is beyond the scope of this paper and we should just mention this in the discussion briefly (I've commented out this section for now).

So, concretely, we'd then have a new appendix which has the following stuff in it:

• Bias-variance tradeoffs
• Complexity--computation trade-offs
• Estimation issues when fitting both global and groupwise smooths
• Alternative formulations: bam, gamm, gamm4, brms (maybe also mention other stuff here too?)

The rest of the content will get cut and some absorbed into other sections. That would then smooth (pun intended) the path for readers into the examples section.

I've not made these more drastic edits for now but can do if we all agree they are okay.

[eric-pedersen]

Would it make sense, rather than appendicizing a lot of this, moving this whole section to after the examples? Right now it almost functions as a discussion / further research jumping off point anyway, and I'd like to avoid having an appendix besides the code supplemental if we can. Since it's a tutorial paper, I think any advice we want to share on how to fit these models should go in the main paper (no one is likely to read an appendix).

[dill]

Yeah okay, let's see what that looks like. I wasn't 100% sure on how many words we're at now but if it fits then that's fine by me.

[eric-pedersen]

@dill, are you happy with the current arrangement? If so, we should be able to close this.

[dill]

:+1: