Gavin Simpson

It is desirable and possible. I've been doing some work under the hood to make it easier to accommodate more models and smooth types, and now that most of that...

Worm plots are now implemented in `worm_plot()` (as of versions > 0.5.1), but not yet active as an option in `appraise()`.

Thanks @Aariq — yes, it fails here because I didn't anticipate (in the code) that you'd have smooths of different dimension. The solution is to do as `smooth_estimates()` and `compare_smooths()`...

_From @alexpghayes on June 15, 2018 17:31_ Seconded. New to GAMs here so I suspect I'm missing something, but is it currently possible to get pointwise prediction intervals for new...

_From @scottkosty on June 15, 2018 19:13_ > Seconded. New to GAMs here so I suspect I'm missing something, but is it currently possible to get pointwise prediction intervals for...

I plan on allowing this in general — see #86 — although I would have to think on the specific case of `confint()`. It won't be via users' own posterior...

I'm going to close this here, but perhaps follow up in general in #86 ?

I see what you mean now and this is different from #86 . So let's reopen this and keep thinking about it. An option might be to allow to provide...

These values are correct; the `est` column is the value of the smooth, which is centred about 0. To recover the values in the paper you need to add on...

As mentioned in the paper, a model with a wiggly trend and a CAR(1) process is often unidentifiable; both components are effectively the same thing if one or the other...