Region covariate

[jpwrobinson]

We probably need to account for species occupying biogeographic regions and so cannot be exported by every country.

For example, in an analysis of species probability of export by country governance levels, we need to constrain cases of countries NOT exporting a species (export = 0), but only if that species is actually found in that country.

Also related issue of rarity when countries fall on range distribution edges. But not sure how we'd tackle that without abundance data.

[jpwrobinson]

Ok, what this should be is 'does the country fall in the species' extent of occurrence?'

We can get this from lat and lon OBIS data in OBIS_dist_species_append.R.

It won't be fun.

[jpwrobinson]

Thinking about this again - I still see it as an unfair test of range size influence on aquarium trade. To take the binomial approach (probability of a country trading a given species), we add zeroes for any species that does not appear in a countries' trade record, so we flood the dataset with zeroes, many of which are not logically true (Jamaica cannot trade an Indo-Pacific species). As a result, we would expect a very strong positive relationship between species trade ~ range size (and this does appear to come out here.

Seeing as this is a pretty obvious pattern, it might be actually better to just account for range size in prediction (i.e. as a fixed covariate), and not have this as one of our focal hypotheses? Species that are heavily traded are likely to have large range sizes, which would confound any trophic patterns..right?

[JamieMcDevittIrwin]

I think that makes sense.. especially if we are exploring right now as a last hurrah. It's easier just to run the models with range size and not have it as one of our hypotheses. If we keep going forward with these hypotheses then we can think about it more and if we think we need to do something else (like your second comment above).