inla.knmodels example crashes

[pav]

I tried to run the example on the help page for inla.knmodels(), and it crashed the INLA system. When running with verbose=TRUE, the final message was:

GMRFLib version 3.0-0-snapshot, has recived error no [3]

Reason    : Matrix is (numerical) singular

Function  : GMRFLib_comp_posdef_inverse

File      : lapack-interface.c

Line      : 71

GitID     : file: lapack-interface.c  1f6a39183ef43d8ef33f10ff3f04fd13f8432758 - Mon Feb 22 21:27:50 2021 +0300

followed by the usual

Error in inla.inlaprogram.has.crashed() :

message. If it would help, I can post the entire verbose message; I don't really understand it.. Given the "experimental" warning about bym2, I tried again with bym and besag, but these crashed INLA too.

Can anyone help?

INLA version  21.02.33

Win10 (64bit)

Thanks!

[pav]

ADDED: actually, "4d" works (with the original bym2), while both "4" and "4c" crash INLA.

[Elias T. Krainski]

There is a typo in the definition of the temporal graph... Please use the following

 tgraph <- sparseMatrix(i=c(2:10, 1:9), j=c(1:9, 2:10), x=1)

Elias

--
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/46b3df7e-15ab-450e-8930-ecb4fce2feb7n%40googlegroups.com.

[pav]

Just to complete the story:

Elias' note that there was a typo in the specification

of tgraph didn't solve the crash problem for me. (I'm

curious: do others experience the inla crash either with

or without the tgraph correction?) In a personal

communication he suggested adding

  control.inla=list(strategy='gaussian')

to the knmodels call. This worked for me (though

I don't understand why).

So a working version of the knmodels help-page

example is:

tgraph <- sparseMatrix(i=c(2:10, 1:9), j=c(1:9, 2:10), x=1)

res <- inla.knmodels(y ~ f(time, model='bym2', graph=tgraph) +

                       f(space, model='bym2', graph=graph),

                     data=dat, family='poisson', E=dat$E, progress=TRUE,

                     control.st=list(time=time, space=space,

                                     spacetime=spacetime, graph=graph,

type=c(4, '4c', '4d')),

                     control.compute=list(dic=TRUE, waic=TRUE, cpo=TRUE),

                     control.inla=list(strategy='gaussian'))

Thanks to Elias Krainski for all his help!