occuMulti Intercepts

[Ettore Fedele]

Hi,

I am using the function occuMulti to fit multi-species occupancy models for two species.

So far I have created two very simple sets of models. The first, keeps both detection and occupancy fixed and equal to 1 for the two species independently while leaving the two-way species interaction at 0. The second is similar to the first, except that the two-way species interaction is now fixed at 1.

My doubt arises when I try to interpret the results, in fact, while for the first model the incerpects are significant, for the second they aren't, not even those for the independent species. I struggle to understand why this is. can anyone help with this?

Thanks,

Ettore

[Ken Kellner]

Hi Ettore,

I am not really sure what you mean by "fixed and equal to 1". I guess you might mean that you modeled occupancy without covariates (intercept-only) using the R formula structure "~1". In this case you are still estimating intercept(s), represented by the 1, your occupancy/detection are not "fixed".

In most cases, there isn't much point in considering the 'significance' of the single-species intercept term(s) even though unmarked does give you a p-value for them. You are essentially testing whether the intercept value is different from 0 on the logit scale, which is 0.5 on the occupancy probability scale. An occupancy value of 0.5 doesn't have any special meaning typically so the test doesn't have any important meaning. So unless you have a specific reason, I would not really pay attention to single-species intercept p-values. 

On the other hand when considering covariates, testing if the covariate parameter value is different from 0 does have an important meaning: 0 would indicate no effect of the covariate, so if the estimate is different from 0 based on the p-value, that covariate is having an effect. This is also true for the interaction intercept: if 0, then you don't have strong evidence that there is an interaction between the species. If different from 0, you do. Thus the p-value has meaning there also.

It is not surprising that both the value and significance would change between intercepts for the two separate species and the intercepts from the multispecies model. These are not measuring exactly the same thing. In the single-species models the intercepts represent the average occupancy for each species, but in the multispecies model the intercepts represent estimates of the "natural parameters" associated with occupancy of each species at a site without the other species present. Additionally, given the added complexity of the multispecies model relative to separate single-species models you may end up with more uncertainty in the estimates of each parameter, which will of course change the associated p-values and make them larger.

Ken