Smooth term with random walk (RW2) in GAM using sdmTMB

[matteba]

Hi,
I'm trying to apply a smoother term for a environmental covariates in a GAM using the fuction model random walk (rw2) as implimented in INLA (see https://becarioprecario.bitbucket.io/inla-gitbook/ch-smoothing.html#sec:smoothterms).
I was wondering if anybody came across this before or knows how it is possible to impliment it on sdmTMB.
Thanks

[seananderson]

I'm pretty sure that's not currently possible. It might be possible if it's possible to include that in an mgcv s() call somehow. sdmTMB will let you have coefficients that follow a random walk or AR(1) process (time_varying formula and time_varying_type arguments) or use smoother functions with s() or t2() (including s(..., by =), k =, 2D/3D smoothers, and setting knots via the knots argument). I haven't carefully followed everything going on with 'rw2' in INLA, but the wiggliness of the smoothers in sdmTMB is penalized via the magic of mgcv::smooth2random.