Error using two natural splines in one formula

[rmip...@gmail.com]

Hi all,

Is there a way around a fault I'm getting when using two natural splines, via the ns function. Here's a toy example:

require(splines) ; require(INLA)

set.seed(12344)

x=rnorm(100)

y=rnorm(100)

z=rnorm(100)

data=data.frame(x=x,y=y,z=z)

fml = y ~ ns(x) + ns(z)

mod = inla(formula = fml,data=data)

When this runs, I get:

 *** Key [1] is used twice and that is not allowed.

 *** A typical example where this happens is: y ~ x + f(x)

 *** Change this formula into: y ~ x + f(x2)

 *** where you define x2=x

it seems to be something to do with the way that the subset of each spline is fed into the INLA formula, but am currently unsure how to fix it.

Any help would be very welcome.

Thanks a lot.

Robbie

[Helpdesk]

INLA does not us 'ns', but its own version of ``splines'', see
inla.doc("rw2")

> --
> 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/b3884f4b-6e95-40ec-a99b-ebf36e6eb66an%40googlegroups.com
> .

--
Håvard Rue
he...@r-inla.org

[rmip...@gmail.com]

Thanks so are rw2 and splines equivalent? 

[INLA help]

Read chapter 3 in the GMRF book 

—

Haavard Rue

HelpDesk 

help@r-inla. org

To view this discussion on the web, visit https://groups.google.com/d/msgid/r-inla-discussion-group/f7a13bec-4b08-4ca5-a0e6-2e9fbc0c9a70n%40googlegroups.com.

[Finn Lindgren]

Hi,

frequentist use of splines and Bayesian latent Gaussian models are never exactly equivalent since one refers to a maximum likelihood estimator and the other refers to a stochastic process model. However, ordinary cubic spline estimates are "equivalent" to the rw2 process model in the sense that for a rw2 process observed with additive independent gaussian noise, the posterior mode is the same the minimiser of a least squares problem with a cubic spline penalty, since the cubic spline minimises the same energy functional; the integral of the square of the second order derivative.

Finn

To view this discussion on the web, visit https://groups.google.com/d/msgid/r-inla-discussion-group/f7a13bec-4b08-4ca5-a0e6-2e9fbc0c9a70n%40googlegroups.com.

--

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

[rmip...@gmail.com]

Thanks that's really useful!