Tasks to do for bbgdm package. - Githubissues

skiptoniam / bbgdm

bbgdm: Assess uncertainty in Generalised Dissimilarity Models via a Bayesian bootstrap

GNU General Public License v2.0

2 stars 2 forks source link

Tasks to do for bbgdm package. #1

Open skiptoniam opened 7 years ago

skiptoniam commented 7 years ago

[x] Just catch bb parameter estimates to reduce model size.
[ ] Fix link function in predict.bbgdm
[ ] Fix i-spline transformation in predict.bbgdm
[ ] Add return eta to predict.bbgm
[ ] Add return whole dissimilarity to predict.bbgdm
[ ] Add a function to incorporate custom dissimilarity
[ ] Extrapolation beyond data?

Leitemfa commented 5 years ago

Is there any interest in including other family functions such as negative binomial?

skiptoniam commented 5 years ago

No. Because we are modelling dissimilarity data, so only Bernoulli, binomial or beta are appropriate.

Leitemfa commented 5 years ago

Hi @skiptoniam , thanks for the quick reply. I was just wondering if there will be a away to analyze dissimilarity from abundance instead only presence absence as it is available now.

Leitemfa commented 5 years ago

Is there any plan to include that?

skiptoniam commented 5 years ago

@Leitemfa usually modelling dissimilarities from abundance data is done via some specific type of dissimilarity metric like bray-curtis (and others). If you want to model multiple species where you have abundance data you might want to consider multiple species models like mvabund, joint species distribution models HMSC or our other work which models multiple species as groups ecomix. Under all these models you can deal with multiple species and abundance data explicitly, it just really depends on what your question is...

Leitemfa commented 5 years ago

Hi @skiptoniam , yes, I understand that, but I see in the bbgdm package that it only analysis presence absence data, but I would like to use it to investigate abundance data, just like in gdm package but including the boostrap framework available in bbgdm. Is that possible?

skiptoniam commented 5 years ago

@Leitemfa yep If you chose the "bray_curtis" option in dissimilarity.

Leitemfa commented 5 years ago

Well, but when I add an abundance data and change the dism_metric to "bray_curtis" I got the message:

Error in pam2dissim(sp.dat, dism_metric) : bbgdm only works with presence-absence - transform your abundances to binary data

skiptoniam commented 5 years ago

Woops. Ok, I can change that. I'll let you know when I have.

Leitemfa commented 5 years ago

Thank you very much @skiptoniam . Looking foward to see that.

Kind regards, Marcio