Open jpwrobinson opened 6 years ago
@jpwrobinson So is our response variable now n. trade countries per species instead of binomial (yes or no its traded)?
So I reran the range model with y=n of exported countries and it does explain a lot more of the variation (Rm=.29) than the binomial response. But that makes sense for the range size hypothesis...
yes! Seems a little more connected to our question I think. We can talk about the binomial approach again Monday maybe
On 31 January 2018 at 18:59, Jamie McDevitt-Irwin notifications@github.com wrote:
@jpwrobinson https://github.com/jpwrobinson So is our response variable now n. trade countries per species instead of binomial (yes or no its traded)?
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/baumlab/open-science-project/issues/14#issuecomment-362034787, or mute the thread https://github.com/notifications/unsubscribe-auth/AE_TKkw7Tnsj2ECCX0yPF6EcEYtoMgcgks5tQLgpgaJpZM4Rw-k2 .
Dr. James Robinson
Senior Research Associate Lancaster Environment Centre Lancaster University, Lancaster, UK
Yeah i think it makes more intuitive sense, but I don't understand how we will have governance as a fixed effect if the y response is the number of countries exporting the species? I guess we could split up models based on governance as a categorical variable? We could also change the response variable for different parts of our question?
Sorry if I'm just really confused here :)
Ah damn you're totally right. Will have a think about that.
On 31 January 2018 at 19:28, Jamie McDevitt-Irwin notifications@github.com wrote:
Yeah i think it makes more intuitive sense, but I don't understand how we will have governance as a fixed effect if the y response is the number of countries exporting the species? I guess we could split up models based on governance as a categorical variable? We could also change the response variable for different parts of our question?
Sorry if I'm just really confused here :)
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/baumlab/open-science-project/issues/14#issuecomment-362043465, or mute the thread https://github.com/notifications/unsubscribe-auth/AE_TKmoudzy10niAdu9Md2qL1lh1pjc9ks5tQL73gaJpZM4Rw-k2 .
Dr. James Robinson
Senior Research Associate Lancaster Environment Centre Lancaster University, Lancaster, UK
For governance we have to use keep country data separate, so maybe run two different models with each response? I'm going to work more on it today.
Other option is maybe a composite measure of governance from PCA or something to get an "average" governance per species over all countries it is found in. No idea if that makes any sense.
I guess we need to decide if species is our unit of interest or country (or both)
I kind of feel like it might be good to do some exploratory plots on both to get an idea for moving forward.. because both are different ways of attempting our question.
I think your idea of a composite measure of a PCA makes sense, but still think it would make more sense to look at governance with the y=binomial exported or not
You said it right I think Geoff. One analysis for species - which species are targeted and why? - and one analysis for countries - how does governance modify which species are targeted?
The species analysis would be number of countries exporting ~ trophic traits + range size.
Country analysis would be export (yes/no) ~ trophic traits + governance metrics + range size. I still think we'd need to exclude species 0s when they cannot actually be caught in one country (i.e. outside their endemic range). Otherwise we will massively inflate the dataset with false 0s.
N countries exporting species increases with area of occurrence (makes sense). After accounting for this, other variables are not good predictors.
Haven't brought governance in here though
We are finding that range size and trophic level are poor predictors of aquarium trade.
Our other hypothesis relates to governance - does high/low marine governance control traded species? Likely then that this also affects range size + trophic level relationships - perhaps low governance countries trade more vulnerable or higher TL species than high governance countries?
Test this by examining n. traded countries (per species) by range size and trophic level across different governance levels (i.e. the interaction).