Skip to main content
×
Home
    • Aa
    • Aa

On a conservative Bayesian method of inferring extinction

  • John Alroy (a1)
Abstract
Abstract

Few methods exist that put a posterior probability on the hypothesis that a thing is gone forever given its sighting history. A recently proposed Bayesian method is highly accurate but aggressive, generating many near-zero or near-one probabilities of extinction. Here I explore a Bayesian method called the agnostic equation that makes radically different assumptions and is much more conservative. The method assumes that the overall prior probability of extinction is 50% and that slices of the prior are exponentially distributed across the time series. The conditional probability of the data given survival is based on a simple, long-known combinatorial expression that captures the chance all presences would fall at random before or within the last sighting’s interval (L) given that they could fall anywhere. The same equation is used to compute the conditional probability given extinction at the start of each interval i, that is, the chance that all sightings would fall before or within L given that none could equal or postdate i. The conditional probability is zero for all hypothesized extinction intervals through L. Bayes’s theorem is then used to compute the posterior extinction probability. It is noted that recycling the mean posterior for a population as a prior improves the method’s accuracy. The agnostic equation differs from an earlier, related one, because it explicitly includes a term to represent the hypothesis of survival, and it therefore does not assume that the species has necessarily gone extinct within the sampling window. Simulations demonstrate that the posterior extinction probabilities are highly accurate when considered as a suite but individually indecisive, rarely approaching one. This property is advantageous whenever inferring extinction would have dangerous consequences. The agnostic method is therefore advocated in cases where conservative estimates are desired.

Copyright
Linked references
Hide All

This list contains references from the content that can be linked to their source. For a full set of references and notes please see the PDF or HTML where available.

B. Collen , A. Purvis , and G. M. Mace . 2010. When is a species really extinct? Testing extinction inference from a sighting record to inform conservation assessment. Diversity and Distributions 16:755764.

Recommend this journal

Email your librarian or administrator to recommend adding this journal to your organisation's collection.

Paleobiology
  • ISSN: 0094-8373
  • EISSN: 1938-5331
  • URL: /core/journals/paleobiology
Please enter your name
Please enter a valid email address
Who would you like to send this to? *
×

Metrics

Altmetric attention score

Full text views

Total number of HTML views: 6
Total number of PDF views: 43 *
Loading metrics...

Abstract views

Total abstract views: 165 *
Loading metrics...

* Views captured on Cambridge Core between September 2016 - 29th April 2017. This data will be updated every 24 hours.