Aki Vehtari comments

Results 437 comments of


                                            Aki Vehtari

A notebook/html has content for many chapters - a marker would be helpful

Thanks for the suggestion. I agree with this and I'll keep this issue open until I have time to add some additional comments to the code.

Figure 16.2

The code for that figure is not in the examples repository and the code is that old that Andrew didn't yet find the code, but could recreate the code if...

Solutions

Hi @LauraRK, sorry we have not made solutions. Do you need clarifications for some of the exercises?

Solutions

Dear @giordafrancis, I hope you don't publish solutions to all exercises as that will reduce the value of the book for those teachers that would like to use the exercises...

Missing xbox survey data

It's not open data, but we are re-checking whether it can be included or add a mention if it can't be included.

Add rstan::extract()-like feature to fit$draws()

> to return the following object: This wish looks like an issue for the posterior package proposing a new helper function

During-sampling diagnostics (feature request & design discussion)

- This looks more like an cmdstan issue. - As ESS and Rhat are computed for scalar variables the computation cost during sampling can be significant for models with a...

Add a vignette example of handling posterior draws for vectors

Link to a draft case study using this for a GP model https://avehtari.github.io/casestudies/Motorcycle/motorcycle.html (I'm using that in teaching next Monday and then I'll remove the bad optimization results)

Variational inference in cmdstanr missing importance_resampling that is in rstan

> If possible maybe what we should do is move most of the code for these features into the loo package itself. `cmdstanr` is using `posterior` package that has support...

Variational inference in cmdstanr missing importance_resampling that is in rstan

> So lp__ would be the log of total probability (including prior) up to an additive constant, right ? Yes, and that is sufficient for the importance sampling used. >...