summary.hyperpar providing erroneous results

Guillermo

unread,

Apr 1, 2021, 11:47:46 AM4/1/21

to R-inla discussion group

Hi,

A few months ago I run INLA, and used pc.priors for the parameters Soak.time (random walk smoother) as seen in the file attached "rw prior". The summary.hyperpar output from this model is shown in the file attached "summary.hyperpar_FIRST"

I have not touched this model in a while, and in the meantime, I have updated INLA in my computer and also my R version.

Recently, I run this model again, using the same pc.prior formulation, as well as the same dataset as the first time (Just wanted to check everything is fine after the updates as I am presenting this work soon)....However, as you can see in the file attached "summary.hyperpar_SECOND", the outputs are very different, particularly in the Soak_Time, fvesselID and the range of w....

What could be causing the issue? Could it be related to the definiton of the priors? Has there been any major change in the past months? Or Am I missing something else?

Also, when running this model now, I am getting the in the INLA verbose the following:

*** WARNING *** Eigenvalue 2 of the Hessian is -14.9846 < 0

*** WARNING *** This have consequence for the accurancy of the hyperpar

*** WARNING *** Continue with a diagonal Hessian.

which I do not recall seeing before

Kind regards

Guillermo

summary.hyperpar_Second.png

rw prior.png

summary.hyperpar_FIRST.png

INLA help

unread,

Apr 1, 2021, 12:15:21 PM4/1/21

to R-inla discussion group, Guillermo

I need to rerun it here to check. Can u send so I can do that?

—

Haavard Rue

HelpDesk

help@r-inla. org

--
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/2c04022c-0003-486d-9ebd-3faccbc68051n%40googlegroups.com.

Guillermo

unread,

Apr 6, 2021, 5:05:51 AM4/6/21

to R-inla discussion group

Dear Haavard,

Thanks for your reply. I will prepare the files and send them to you privately.

In the meantime, I asked a colleague to run the same model on his computer, and he got the same results as I got a year ago.

He has not updated INLA since then, and thus his currently INLA version is: "INLA_19.09.03". My version however is "INLA_21.02.23".