Poorly Fitting Model

[Sylvan Benaksas]

Hi There,

I have created a spatiotemporal tweedie model fitted to fish biomass data, I have created a first model that fits reasonably well, however my second model for a different fish species has a poor fit. Fitted values are far lower than the underlying data (~ x6). In spite of being too small, the shape of most covariate-response relationship fits well except for one variable that has an unexpected fit. I have tried changing all priors but can not improve the fit, can anyone offer some advice what is going wrong here?

See one covariate fit here, observed in black and fitted in red, the relationship looks right but values too small

Now see the misbehaving covariate, the relationship is fit all wrong

I can provide data and code if anyone can check what is the issue?

Kind regards,

Sylvan

[Elias T. Krainski]

Hi,

From my litle experience with this kind of data this kind of result is not rare. 

Since depth is related to spatial location and you have high variability at some depth values I guess that the spatio-temporal effect captured it. 

Since salinity varies over spacetime, maybe there is some interaction with the spatio-temporal effect. 

That being said, it is easier to control this kind of random effects than the fixed ones. I meant you can put stronger prior or even fix the hyperparameters if you want specific behaviour for the related random effect property. 

Best regards, 

Elias

--
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/dc96f0b7-7af8-4c58-9cd3-89dd50327226n%40googlegroups.com.