Re: [r-inla] INLA model comparison (CPO/PIT and/or MSE/MSEP)

[INLA help]

On Sat, 2016-01-16 at 10:15 -0800, Bruno da Costa Perez wrote:
> Hello, 
>
> I'm trying to compare different models in R-INLA.
>
> Considering that DIC has it's limitations on model comparison...


Hi, 

there is also a related one that tends to work better, WAIC,  see the
blog of Andrew Gelman and search for 'WAIC'. 

you can do waic in INLA, with control.compute = list(waic=TRUE)


WAIC is similar to CPO

http://www.stat.columbia.edu/~gelman/research/unpublished/loo_stan.pdf

and this report (with its references) should, I guess, answer your
questions.

Best,
H

> ...two other possibilities come to my mind, which brings me some
> questions:
>
> - Should I use the CPO/PIT results, by using
> control.compute=list(cpo=TRUE) ?
> And how should I compare those values between models in order to
> select the best?
>
> - Have anyone performed any mean squared error and/or mean squared
> error of prediction (upon cross-validation)?
> Is there any restriction for INLA models to be compared by these
> methods?
>
> And, is model$summary.fitted.values the best way to obtain predicted
> values (in the reponse variable scale) from a model?
>
> Thank you very much in advance,
>
> Kind Regards
>
> Bruno
> -- 
> 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 post to this group, send email to r-inla-discussion-group@googlegr
> oups.com.
> Visit this group at https://groups.google.com/group/r-inla-discussion
> -group.
> For more options, visit https://groups.google.com/d/optout.

--
Håvard Rue
he...@r-inla.org

[Bruno da Costa Perez]

Thank you very much for the reply, Dr. Rue.

Awsome material from Andrew Gelman.

> send an email to r-inla-discussion-group+unsub...@googlegroups.com