INLA.Data1: Gaussian weight[13361] = 0 is void

[Doori Oh]

Hi everyone,

I'm getting a new error using the same environments, configurations, and datasets which successfully ran before. I will attach the errors I received. It's odd because I don't have 0 in weight column. I use family = "gaussian". I would greatly appreciate any help/suggestions!

*** ERROR ***     INLA.Data1: Gaussian weight[13361] = 0 is void

Error in inla.inlaprogram.has.crashed() : 
  The inla-program exited with an error. Unless you interupted it yourself, please rerun with verbose=TRUE and check the output carefully.
  If this does not help, please contact the developers at <he...@r-inla.org>.
Error in inla.core.safe(formula = formula, family = family, contrasts = contrasts,  : 
  *** Fail to get good enough initial values. Maybe it is due to something else.
Calls: fit_mbg -> system.time -> inla -> inla.core.safe
Timing stopped at: 3.35 5.318 15.05
Execution halted

[Helpdesk (Haavard Rue)]

I guess you run the most recent testing-version, if you please upgrade.

what happen is that option 'scale' has element that is 0, like

> r=inla(y ~ 1, data=data.frame(y=1:2), scale=c(1,0), safe=FALSE)


*** ERROR *** INLA.Data1: Gaussian weight[1] = 0 is void



saying that the precision of for y[2] is 0, which is not allowed.

if you think this is an error, I need to rerun your code/data here...

H

> --
> 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/7f097789-324a-4750-8d8b-e21b4643b7fan%40googlegroups.com
> .

--
Håvard Rue
he...@r-inla.org

[Dany]

Hi Havard,

I've been using the most latest version (not the testing version). Unfortunately, data are all credentials and I cannot share it. I didn't specify 'scale' in my code and am trying to understand how scale works from your example. Can you let me know the relevant resources or how to test it using my datasets on my side? I'm lost. Thank you. 

[Helpdesk (Haavard Rue)]

Gaussian likelihood for the data y_i has precision \tau, say, which is
the precision parameter in the likelihood,

with scale=c(1,2,3)

then y_1 has precision 1*\tau
y_2 has precision 2*\tau
y_3 has precision 3*\tau

see inla.doc("gaussian")