e=0 ?

[Gaby]

I'm a beginner BioGeoBEARS user running a BSM (1000 replicates) with the DEC+J model on a dataset of 21 taxa and 9 areas (max_range_size=5). The analysis runs fine, but I'm confused by the results.

The output summary table shows both the a and e parameters as 0. As I understad, the DEC model does not include the "a" parameter, so a=0 would make sense. However, I expected the extinction rate e to be non-zero. Is it plausible for the ML estimate of e to be exactly zero, or could this indicate an issue with my dataset or model setup?

As I said, everithing else looks normal and the script ran smoothly without errors.

Any thoughts on what might be happening would be very helpful.

Thanks for your help!

[Ivan Magalhães]

Dear Gaby

The DEC model implemented by BioGeoBEARS has long been known to underestimate extinction/extirpation rates, frequently estimating the rate to be 0, even in simulated scenarios where it is known to be higher. See the original paper describing DEC, Fig 1B: https://academic.oup.com/sysbio/article-abstract/57/1/4/1703014?redirectedFrom=fulltext

So, this is to be expected...

Best,

Ivan

--
Dr. Ivan L. F. Magalhaes

Investigador Asistente  — CONICET
Curador Asociado  —  Colección Nacional de Aracnología
Zootaxa Associate Editor — Araneae: Filistatidae, Synspermiata (in part)

Docente  —  Programa de Pós-Graduação em Biodiversidade Neotropical, UNILA

División Aracnología

Museo Argentino de Ciencias Naturales "Bernardino Rivadavia" 
Av. Ángel Gallardo 470, C1405DJR, Buenos Aires, Argentina

+54 11 4982-8370 ext. 169

Website | ResearchGate | Twitter |

iNaturalist | Lattes | CV-Ar | ORCID | GitHub

--
You received this message because you are subscribed to the Google Groups "BioGeoBEARS" group.
To unsubscribe from this group and stop receiving emails from it, send an email to biogeobears...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/biogeobears/1b5dcc57-04fd-4cbf-b65c-2bca9a693eean%40googlegroups.com.

[Nick Matzke]

Yes, that's right - e is often estimated to be 0, except in BAYAREALIKE models - it competes with processes like vicariance, which produce smaller range sizes via speciation.  BAYAREALIKE does not change the range at speciation, so it's forced to use both range expansion (d) and range contraction (e) to get species around to the different tip ranges.

Cheers,

Nick

To view this discussion visit https://groups.google.com/d/msgid/biogeobears/CAAsMhYo_nNX2O%3D8BhZVOnbuiwwZ052ZkAKt23QKWWVpq_Oz19Q%40mail.gmail.com.