Nicholas Clark

The {`mgcv`} equivalent is very useful, though not widely known about: https://www.rdocumentation.org/packages/mgcv/versions/1.9-1/topics/mgcv.FAQ. Things to cover 1. How to compare models using `loo_compare()` and `lfo_cv()` ([`loo` less reliable for models with...

Would likely require methods for: - [ ] `vcov()` with same arguments as `vcov.gam()` - [ ] update `get_vcov()` to return the Bayesian covariance matrix; shouldn't impact `marginaleffects` functionality -...

Allow hierarchical or no pooling of trend components

1

Right now autoregressive and trend variance parameters are 'hierarchical', but the hyperparameters are are fixed (i.e. `ar1 ~ normal(0, 0.5)`). It would be useful to allow options to learn these...

Add support for `E_loo()`

1

PSIS-LOO can be used to generated weighted expectations and predictions, valuable for asking what predictions we would get if high-leverage "outliers" were downweighted: https://paul-buerkner.github.io/brms/reference/loo_predict.brmsfit.html.

Future enhancement wishlist

6

- Matern covariance functions for GP effects / trends - 2d FFT GPs (or even 1d) for complex but stationary GP effects (https://arxiv.org/pdf/2301.08836.pdf) - Multivariate normal observation models - Move...

Allow bounds on fixed effect priors

5

Incorporating bounds on fixed effect priors can be done using the truncation syntax i.e. ```b[1] ~ normal(0, 1)T[0, ]``` This would require checks on inits to ensure they respect the...

Various areas throughout the package will fail if a matrix with only one row or column is subset (very annoying!). Need to use `drop = FALSE`

Many time series show multiplicative effects, such as seasonality that changes as the underlying level changes. These are easily captured in GAMs using a tensor product (i.e. `te(time, season)`), but...

All fitted models should have a non-null `priors` slot with whatever prior distributions were used. This will make it easy to compare prior and posterior draws, for example

In `mgcv`, the predict function allows certain terms to be omitted. Should be able to include this in `mvgam` by finding the coefficients associated with excluded terms and setting their...