Anisotropy possibility in R-INLA SPDE

[Ruby Ji]

Dear R-inla group,

I am checking to see if there have been any updates on setting anisotropy in INLA-SPDE. I inquired with your team a few years ago, and there were no such settings available at that time. There are some packages inspired by R-INLA, but based on the TMB approach, that include an anisotropy option. Typically, these packages add a rotation matrix, H, to the range parameter. In my opinion, it is similar to the Barrier model within the INLA framework (when I say "similar", I mean they both adjust the range parameter, but of course, differently).

I saw a post from Finn last year mentioning that your team is working on a better approach to address the anisotropy issue. I would like to get the most recent updates on this.

Thanks,

Rujia

[Helpdesk (Haavard Rue)]

I know Finn Lindgren and a student has worked on it, if its there yet in
inlabru(), I don't know. Its just a matter of coding it up. (you can do that
using 'rgeneric' if you do that yourself)

> --
> You received this message because you are subscribed to the Google Groups "R-
> inla discussion group" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to r-inla-discussion...@googlegroups.com.
> To view this discussion on the web, visit
> https://groups.google.com/d/msgid/r-inla-discussion-group/d06fc9a3-d3ea-42f3-ad9e-62ef004b5cfcn%40googlegroups.com
> .

--
Håvard Rue
he...@r-inla.org

[Finn Lindgren]

Hi,

my student has a non-inla implementation for the experiments in the
paper he's writing, but no inla-friendly version yet.

Finn

> You received this message because you are subscribed to the Google Groups "R-inla discussion group" group.

> To unsubscribe from this group and stop receiving emails from it, send an email to r-inla-discussion...@googlegroups.com.

> To view this discussion on the web, visit https://groups.google.com/d/msgid/r-inla-discussion-group/461a34ecc5265bf48285f2589e6a0b744703dd20.camel%40r-inla.org.



--
Finn Lindgren
email: finn.l...@gmail.com

[Ruby Ji]

I have been away from INLA for some time and wasn't aware of 'rgeneric.' Thank you for letting me know. I will start working on it.  

[Berin Mackenzie]

Dear R-INLA group,
I'm just wondering if there have been any further developments regarding implementing anisotropy in INLA over the past year?

Warm regards,

Berin

[Helpdesk (Haavard Rue)]

its 'just' a matter of coding and finding the good parameterisation (which Finn
has solved now). I'm not sure how far Finn has taken this, I guess he will
update on this issue here



On Tue, 2025-06-03 at 01:17 -0700, 'Berin Mackenzie' via R-inla discussion group
wrote:

[Finn Lindgren]

We've been focusing on the theory for how to construct sensible priors, and haven't had time to do coding for inla yet.

Our submitted paper about anisotropy parameterisation and prior construction (in the stationary case) is at

  https://arxiv.org/abs/2409.02331

There may be a basic implementation of an older anisotropy parameterisation in the rSPDE package, but we haven't implemented the new model yet.

Finn

To view this discussion, visit https://groups.google.com/d/msgid/r-inla-discussion-group/c50d321f5cd53ef1f71175f6fbeb13aab3ed618d.camel%40r-inla.org.