Requested update to the 'cenpoisson' likelihood

[joenomiddl...@gmail.com]

Hi INLA team!

I am wanting to fit a censored Poisson model to my data. I have a combination of K different measuring instruments that each have different (but known) censorship intervals (e.g. instrument 1 cannot record the exact counts ( call it N_i ) when N_i >= 10, instrument 2 cannot record exact counts when N_i >=20, etc., ). 

So far it appears that the 'cenpoisson' only accepts a single interval of length 2 as a vector through the argument: 

control.family = list(cenpoisson.I = interval).

Instead, with n datapoints, I would ideally like to define an n x 2 matrix of intervals, with the first column equal to the L values (as defined in the document inla.doc('cenpoisson')) and the second column equal to the H values.

This would save me having to define a joint model with K inla.stack objects, where K is the number of measuring instruments. My value of K is very large (in the hundreds).

After looking at the inla.c source code, I think this should be possible to implement in the function 'loglikelihood_cenpoisson'. Do you think you would be able to make these changes to the cenpoisson likelihood? 

Thanks for providing such an awesome software tool!

Best,

Joe

[Helpdesk]

a very peculiar problem...

yes, allthough its doable with the current code, its messy little
efficient.

I'll see what I can do

> --
> 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/cc1a50c3-35e8-4234-ac83-16cef10dde53n%40googlegroups.com
> .

--
Håvard Rue
he...@r-inla.org

[joenomiddl...@gmail.com]

Thanks, Håvard!

I agree - it's not a typical problem... :P

Best,

Joe