Interpreting categorical RSF output

[Lara Howard]

Hi Chris (and the ctmm group),

I was wondering if anyone could help me with some issues with RSF interpretation for a model looking at categorical variable effects. I can only really find information on interpreting continuous models, so if anyone could offer some resources/papers that have detail on interpreting categorical models, that would be much appreciated. 

I am running an RSF to try to look at broad scale landscape use, looking at range distribution and if habitat can predict this. I am using GPS data from a population of Crowned Eagles and a national landcover dataset, which categorises the country into one of nine different landcover classes.

I have ran a preliminary RSF using rsf.fit with the Riemann integrator, the output of which I have attached below though I'm not entirely sure how to interpret it. From what I understand, category 4 (which is forested land) is being used as a reference category with respect to all other levels, with the insignificant results having much more variance. Forested land is the most used land class, so I appreciate it's probably not the best one to use in the final model. I was thinking about if it's possible to model without a reference category, and wondered if I created binary rasters of each habitat class (1 for habitat class presence and 0 for absence) I could assess each class independently? Would this be appropriate for the model/ do you have any further suggestions?

Happy to attach the code too if that will help. Thanks so much in advance. 

Best wishes,

Lara

[Christen Fleming]

Hi Lara,

It is mathematically impossible to not have a reference category, because the 8 categories are linearly dependent - if you aren't in categories 1-8, then you must be in category 9. When you fit to perfectly co-linear predictors like this, you get numerical errors. You can choose your reference category, though. The default is the most used category.

Best,

Chris