Capturing vb.correction aborted output

[Eric Weine]

Hello,

I'm fitting INLA models to some very sparse spatial transcriptomics data with a Poisson likelihood. I'm fitting the same model for 100s of different genes, so I need to run this on a server and later look at the results.

When fitting the model to some genes, I sometimes get the error "'vb.correction' is aborted". For the genes where this is occurring, I would like to further investigate the model and see if I can find any issues with the data / model assumptions. However, I cannot figure out how to actually capture when this is happening from the R output.

If I run 

fit <- inla(..., verbose = FALSE)

and then examine fit$logfile, the 'vb.correction' message is not stored even though it is being printed in the console.

I tried just saving all console output by running sink() before running inla with verbose = TRUE, but that doesn't seem to capture any INLA output.

Any advice on how I can capture this?

Thanks,

Eric.

[Helpdesk (Haavard Rue)]

Thank you Eric, you're right. I have fixed the code. in the next testing version
you'll see this msg's also in the logfile.

meanwhile, you can do a workaround like in the attached script to check.
hopefully the new testing version should be out next week.


PS: the issue is likely due to vague priors and maybe an excess of zero's of
your poisson data. consider using more informative priors.

Best
Havard

> --
> 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, visit
> https://groups.google.com/d/msgid/r-inla-discussion-group/70016d12-cb75-41c1-9e33-691c51f748een%40googlegroups.com
> .

--
Håvard Rue
he...@r-inla.org

[Helpdesk (Haavard Rue)]

updated script with comments that make sense ;-)

--
Håvard Rue
he...@r-inla.org